- Download PDF-

SEEFOR 8 (2): 71-84
Article ID: 158
DOI: https://doi.org/10.15177/seefor.17-17

Original scientific paper

 

Biogeochemical Modelling vs. Tree-Ring Measurements - Comparison of Growth Dynamic Estimates at Two Distinct Oak Forests in Croatia


Maša Zorana Ostrogović Sever1, Elvis Paladinić1, Zoltán Barcza2,3,4, Dóra Hidy5, Anikó Kern6, Mislav Anić1, Hrvoje Marjanović7*


(1) Croatian Forest Research Institute, Division for Forest Management and Forestry Economics, Trnjanska cesta 35, HR-10000 Zagreb, Croatia;
(2) Eötvös Loránd University, Department of Meteorology, Pázmány P. st. 1/A, H-1117 Budapest, Hungary;
(3) Eötvös Loránd University, Faculty of Sciences, Excellence Center, H-2462 Martonvásár, Brunszvik u. 2., Hungary; 
(4) Czech University of Life Sciences Prague, Faculty of Forestry and Wood Sciences, Kamýcká 129, 165 21 Prague 6, Czech Republic;
(5) Szent István University, MTA-SZIE Plant Ecology Research Group, Páter K. u.1., H-2103 Gödöllő, Hungary;
(6) Eötvös Loránd University, Department of Geophysics and Space Science, Pázmány P. st. 1/A, H-1117 Budapest, Hungary;
(7) Croatian Forest Research Institute, Division for Forest Management and Forestry Economics, Cvjetno naselje 41, HR-10450 Jastrebarsko, Croatia

* Correspondence: e-mail:

Citation: OSTROGOVIĆ SEVER MZ, PALADINIĆ E, BARCZA Z, HIDY D, KERN A, ANIĆ M, MARJANOVIĆ H 2017 Biogeochemical Modelling vs. Tree-Ring Measurements - Comparison of Growth Dynamic Estimates at Two Distinct Oak Forests in Croatia. South-east Eur for 8 (2): 71-84. DOI: https://doi.org/10.15177/seefor.17-17

Received: 28 Jun 2017; Revised: 29 Oct 2017; 29 Nov 2017; 12 Dec 2017; Accepted: 14 Dec 2017; Published online: 21 Dec 2017


Cited by:     Crossref  (1)     Google Scholar


Abstract

Background and Purpose: Biogeochemical process‑based models use a mathematical representation of physical processes with the aim of simulating and predicting past or future state of ecosystems (e.g. forests). Such models, usually executed as computer programs, rely on environmental variables as drivers, hence they can be used in studies of expected changes in environmental conditions. Process‑based models are continuously developed and improved with new scientific findings and newly available datasets. In the case of forests, long-term tree chronologies, either from monitoring or from tree-ring data, offer valuable means for testing modelling results. Information from different tree cores can cover a wide range of ecological and meteorological conditions and as such provide satisfactory temporal and spatial resolution to be used for model testing and improvement.
Materials and Methods:
In our research, we used tree-ring data as a ground truth to test the performance of Biome-BGCMuSo (BBGCMuSo) model in two distinct pedunculate oak forest areas, Kupa River Basin (called Pokupsko Basin) and Spačva River Basin, corresponding to a wetter and a drier site, respectively. Comparison of growth estimates from two different data sources was performed by estimating the dynamics of standardized basal area increment (BAI) from tree-ring data and standardized net primary productivity of stem wood (NPPw) from BBGCMuSo model. The estimated growth dynamics during 2000-2014 were discussed regarding the site-specific conditions and the observed meteorology.
Results: The results showed similar growth dynamic obtained from the model at both investigated locations, although growth estimates from tree-ring data revealed differences between wetter and drier environment. This indicates higher model sensitivity to meteorology (positive temperature anomalies and negative precipitation anomalies during vegetation period) than to site-specific conditions (groundwater, soil type). At both locations, Pokupsko and Spačva, BBGCMuSo showed poor predictive power in capturing the dynamics obtained from tree‑ring data.
Conclusions: BBGCMuSo model, similar to other process-based models, is primarily driven by meteorology, although site-specific conditions are an important factor affecting lowland oak forests’ growth dynamics. When possible, groundwater information should be included in the modelling of lowland oak forests in order to obtain better predictions. The observed discrepancies between measured and modelled data indicate that fixed carbon allocation, currently implemented in the model, fails in predicting growth dynamics of NPP. Dynamic carbon allocation routine should be implemented in the model to better capture tree stress response and growth dynamics.

Keywords: Pedunculate oak forest, basal area increment, net primary productivity, model testing 



INTRODUCTION

In a changing environment, there is a growing need for estimating future forest productivity in order to forecast the impact of climate change on sustainability and adaptability of forests. Process-based modelling is a state-of-the-art technique used in predicting behaviour and future state of ecosystems with respect to environmental conditions [1, 2]. A variety of known ecophysiological and geochemical processes are implemented in these models, but continuous model development based on new knowledge is still needed [3, 4].

Process models, used for vegetation modelling, are complex and often have a high number of driving variables and parameters. This makes calibration (i.e. parameterisation) and validation of such complex vegetation models a challenging task [5]. Model calibration requires an extensive dataset of field measurements, as well as high computational skills and computing power. Most valuable source of field data, used in calibration and validation of vegetation process models, is high‑frequency (i.e. half-hour) eddy-covariance (EC) data [6]. A global network of EC flux measurement sites, such as FLUXNET [7], has great potential in facilitating means for better understanding of carbon dynamics in various biomes across regional and global scales [8]. However, for a particular or specific ecosystem, such as lowland forests of pedunculate oak, this dataset has limited use due to a relatively scarce spatial distribution of flux towers. In addition, even if flux measurements for the selected forest type do exist, single site measurements cover a relatively small area (few hundred meters to few kilometres). Taller towers are capable of covering even larger areas, such as the 82 m high tower in Hegyhátsal, Hungary, but in that case, fluxes reflect a multitude of different land covers [9].

Other sources of data that might be useful in assessing the results of modelling with process‑based models are long-term chronologies from monitoring or from tree-ring data. Databases of tree ring measurements usually cover a wide range of ecological and meteorological conditions. As such, these data contain information of satisfactory temporal and spatial frequency to be used for testing performance of complex process-based models. New knowledge gained through model comparison with various measurement datasets is a valuable source of information to be used for model improvements and further model developments [7, 10].

The dendrochronological approach provides a unique long-term understanding of the interplay between terrestrial ecosystems and external forcing agents [11, 12]. It is most suitable for trees of the temperate climate zone. Tree-ring width (TRW) data and its derived variables (e.g. tree basal area increment, BAI), reflect tree’s radial growth due to cambial activity. Tree’s stem growth at a given year often integrates the meteorological effects of the current year and several previous years, and it is further modified by site-specific conditions and management [13]. In this way, in their annual rings, trees preserve an archive of past growing conditions reflecting climate anomalies, competition, disturbance, soil characteristics or species-specific growth patterns [14, 15], as well as human-induced disturbances. Therefore, when using or interpreting tree-ring width data all these influencing factors should be kept in mind.

According to Hafner et al. [16], when analysing the response of lowland oak forests to climate conditions, it is important to consider the micro-environment (e.g. drier vs. wetter sites), but also to distinguish tree-ring data into early- and latewood formations. The underlying idea behind this research was to test modelling performance by using tree-ring data as ground truth, rather than to analyse the climatic influence on tree-ring formations. Therefore, we used a whole tree‑ring width as a proxy for realised annual growth to test the modelled growth dynamics at different locations. Simple visual interpretation of tree growth response to the observed meteorology was performed only with the purpose of providing additional insight into differences between investigated locations which could further be used in defining potential issues in the model logic.

In our research, TRW data, combined with dendrometric data (i.e. diameter at breast height), were used to assess the inter-annual variability of productivity in lowland oak forests. The aim of this study is to test modelling performance by comparing forest growth dynamics estimates from Biome-BGCMuSo model (BBGCMuSo) against the observed growth estimated from an extensive dataset of tree-rings. The observed differences will serve for defining modelling issues and indicating potential directions for further model improvements. There is evidence of growth decrease of pedunculate oak forests in Southeast Europe as a response to a change of water regime and climate [17]. Reliable model predictions are needed for the selection of appropriate adaptation measures for the preservation of those forests.

 

MATERIALS AND METHODS

Study Areas

The research was conducted in two distinct areas of managed pedunculate oak forest in Croatia, Kupa River Basin (also called Pokupsko Basin) located in western part of Croatia, approximately 35 km SW of Zagreb, and Spačva Basin located in eastern part of Croatia, approximately 35 km SE of Vinkovci (Figure 1).

 

FIGURE 1. Geographical location of the research areas, Pokupsko Basin (west) and Spačva Basin (east), and the meteorological stations located in Jastrebarsko and Gradište.

 

The dominant tree species in both forest complexes is pedunculate oak (Quercus robur L.) with a significant share of common hornbeam (Carpinus betulus L.), and narrow-leaved ash (Fraxinus angustifolia Vahl.). Black alder (Alnus glutinosa (L.) Geartn.) is also present, but more abundantly in Pokupsko Basin where it holds a significant share in stock (Table 1). Oak forests in Croatia are managed as even-aged, with 140 years long rotations that end with two or three regeneration cuts during last 10 years of the rotation.

 

TABLE 1. Site description

 

Floodplain forests of Pokupsko Basin grow in the tectonic basin “Crna Mlaka” surrounded by hilly slopes of Samobor, Žumberak and Vukomerec hills on east, west and north side, and Kupa River on the south. The Basin lies between 15°32’ and 15°50’ longitude east, and 45°30’ to 45°42’ latitude north occupying mostly flat area, with altitude ranging from 107 to 115 m a.s.l. The climate in Pokupsko Basin is warm temperate with a mean annual air temperature of 10.6°C and precipitation of 962 mm·y-1 for the period 1981-2010 (data obtained from national Meteorological and Hydrological Service for nearest meteorological station in Jastrebarsko, 45°40’N, 15°39’E, 140 m a.s.l., approx. 2 km NW of the Pokupsko Basin forest). Soils are hydromorphic on clay parent material and according to the World Reference Base for Soil Resources [18], they are classified as luvic stagnosol (Table 1). Average groundwater table depth (based on the data from previous measurements until 1997 and those from 2008 onwards, made by the researchers of Croatian Forest Research Institute), is from 60 to 200 cm [19].

The forest complex of Spačva Basin lays at the most eastern part of Croatia, between Sava and Drava rivers, on the catchment area of Bosut River and its tributaries. Located between 18°45’ and 19°10’ longitude east and 44°51’ to 45°09’ latitude north, it occupies flat-curly basin of altitude ranging from 77 to 90 m a.s.l., which is intersected by numerous small rivers. According to Seletković [20], the climate in the eastern part of pedunculate oak distribution area in Croatia is warm temperate with maximum rainfall in June, without exceptionally dry months in summer, and with driest months occurring during cold period of the year. The mean annual air temperature is 11.5°C and precipitation is 686 mm·y-1 for the period 1981-2010 (data obtained from National Meteorological and Hydrological Service for nearest meteorological station Gradište, 45°10’N, 18°42’E, 89 m a.s.l., approx. located 4 km W of the Spačva forest). Majority of Spačva Basin forest soils are semi-terrestrial or hydromorphic soils on loamy-clay river sediments [21], and according to World Reference Base for Soil Resources [18], they are classified as chernozem. In the period from 1996 to 2012, an average observed groundwater table depth was ranging from 139 to 617 cm [22]. Differences between two forest complexes are summarized in Table 1.

Biome-BGCMuSo Model

Biome-BGCMuSo (BBGCMuSo) [23, 24] is a newer version of the original biogeochemical model Biome-BGC that simulates carbon, nitrogen, and water cycling in different terrestrial ecosystems [25]. In general, Biome-BGC is a process-based model widely used for estimating ecosystem productivity under current and changed environmental conditions [23, 26-30]. Major improvements in BBGCMuSo include introducing a multilayer soil module with the possibility of using groundwater table information, management module, new plant pools, respiration acclimation, CO2 regulation of stomatal conductance and transient run [23, 24].

There are two obligatory input datasets for running the model, namely meteorology and ecophysiological traits of the specific ecosystem, and several optional datasets, e.g. atmospheric CO2 concentration, nitrogen deposition, management data, groundwater table etc. Model simulation has three steps: spin-up, transient run and normal run. The purpose of spin-up is to bring the ecosystem to the steady state regarding soil carbon stocks using long-term local meteorological data. Transient run enables a smooth transition from spin-up phase to the normal phase as it slowly brings the ecosystem to steady state under current (changed) environmental conditions using varying data on CO2 concentration, nitrogen deposition and management. Finally, the normal run is done using current meteorology, CO2 concentration, nitrogen deposition, and management for the period of interest.

In this research, we simulated the productivity of selected stands in managed oak forests in two selected areas, Pokupsko Basin (4 forest management units with 947 forest compartments covering 11.1 kha in total) and Spačva Basin (13 forest management units with a total of 2918 forest compartments covering 47.8 kha in total). Only forest compartments categorised, according to the dominant tree species, in forest management plans as management class of pedunculate oak, older than 15 years in the year 2000, and for which regeneration harvests have not occurred between 2000 and 2014 were considered for simulation and tree coring (potentially 6.4 kha in 524 compartments in Pokupsko and 33.2 kha in 2083 compartments in Spačva Basin). For the selected 2607 forest compartments in both areas, we run the simulations. However, in the comparison of modelled and measured results only those forest compartments where we actually performed measurements were considered (36 in Pokupsko and 55 in Spačva Basin). More details on the selection of forest compartments are given in the next chapter. 

Model simulation was performed on a forest stand level since each forest compartment corresponds to a single forest stand. To account for different management history of stands of different age (i.e. management compartments; for age class distribution see Figure 2), we set the spin-up and transient simulations to the period 1850-1999, while the normal run was done for the period of interest, i.e. 2000-2014. From the records in forest management plans for the year and volume of thinning / regeneration harvests, and the growing stocks at the forest compartment level we calculated the intensities. Specific intensity values were used for the simulation of each forest compartment in the normal run. For the spin-up and transient run we calculated the average intensity of thinning and regeneration harvests and applied those values to all forest compartments. However, the timing (i.e. the year) of the thinning was estimated from the existing records of stand age and year of thinning in the 10-year steps (e.g. thinning in 2009 in the records implies thinning in 1999, 1989, etc., until the final harvest of the previous stand).

 

FIGURE 2. Distribution of the number of model runs (simulations) and the number of tree cores according to age classes for forests of Pokupsko and Spačva basins. Note that each simulation is made at the forest compartment level.

 

Meteorological data used in the simulation was obtained from FORESEE database [31]. FORESEE is a gridded database with a spatial resolution of 1/6o x 1/6o  containing daily maximum/minimum temperature and precipitation fields for Central Europe. In addition, the FORESEE retrieval site (http://nimbus.elte.hu/FORESEE/map_query/index.html) offers the possibility for retrieval of core meteorological variables needed for running Biome-BGCMuSo, namely: the daily minimum, maximum and average daytime (from sunrise to sunset) temperature (°C), daily total precipitation (cm), daylight average vapour pressure deficit (Pa), shortwave radiant flux density (W·m-2) and day length from sunrise to sunset (seconds). By overlapping FORESEE database over a spatial distribution of the selected management compartments, a specific meteorological dataset was assigned to each forest compartment.Considering that this dataset covers the period from 1951, also taking into account that the current minimum prescribed rotation length for pedunculate oak is 140 years, we needed to approximate to the meteorology from 1851. For the purpose of our simulations, we assumed that meteorology for 1851-1950 was the same as it was during 1951-1970. Therefore, we used multiple times data from the period 1951-1970 without randomization. For ecophysiological traits, we used a parameter list for oak forests published in Hidy et al. [24], slightly adjusted to site-specific conditions (Table 2). The main difference between two investigated sites is a share of black alder. Black alder is a nitrogen-fixing species, and therefore a higher nitrogen fixation rate [32], relative to the share of black alder [33]), is used at Pokupsko site. Values of other adjusted parameters are set to default [24] (Table 2). Spatially explicit data on stand elevation was obtained from Croatian Forests Ltd. database containing all information on forest stands, as prescribed by the national regulation [34], while for stand latitude, we used the latitude of the corresponding FORESEE pixel [31, 35]. Site‑specific soil texture was calculated from previously collected soil data, resulting with one texture that was used in simulations at Pokupsko and the other at Spačva Basin [19, 36].

 

TABLE 2. Parameter values adjusted to site-specific conditions.

 

Furthermore, we used three optional input files: atmospheric CO2, nitrogen deposition, and management file. Atmospheric CO2 concentration data were obtained from Mauna Loa Observatory (available online at http://www.esrl.noaa.gov/gmd/obop/mlo/) and using relevant publications [37]. Nitrogen deposition data was based on Churkina et al. [38]. Management data (stand age, wood volume stocks by tree species, the volume of wood extracted with thinning or stand regeneration and year of the activities, soil type, etc.) was obtained from Croatian Forests Ltd. database. For transient run, management was reconstructed using available information on the stand age in each management compartment. The final cut year was set as the first year of stand development, and from that year onward thinning events were set at every 10 years using average thinning rate of 15%. For the normal run, the actual thinning rates were used for each forest compartment. Thinning rates were estimated from records of standing wood stock, estimated annual wood increment, year of thinning and volume of extracted wood available from the Croatian Forests Ltd. database. In order to use groundwater table information, the user should provide daily data. Unfortunately, daily data on groundwater table was not available for both investigated sites; therefore, this model feature was not used.

Tree-Ring Data

A field survey was conducted from spring 2015 to spring 2016. This research was part of the project EFFEctivity (http://www.sumins.hr/en/projekti/effectivity/) which had, in addition to the work presented here, also the goal of testing MODIS MOD17 annual Net Primary Productivity product [39]. This determined the design for the selection of the plot locations. In short, both forest areas, Pokupsko and Spačva, were overlaid with grid corresponding to MODIS 1 km resolution pixels. Only pixels with more than 90% forest cover and with homogenous age structure (forest compartments consisting of >70% of the pixel area had to have the stand age difference of less than 40 years) were selected and in each pixel four plots were installed. The location of the plot was at the centre of the MODIS pixels with 500 m resolution (each 1 km MODIS pixel can be subdivided into four 500 m pixels). The example of the plot layout is presented in Figure 3. In total, 109 temporary circular plots were placed within two investigated forest areas (41 plots in Pokupsko and 68 in Spačva). Sampling radius varied depending on the stand age and tree size of the sampled tree with larger trees being sampled using larger radius [40] (Table 3). On each plot diameter at breast height (dbh, 1.30 m above the ground) of all sampled trees, as well as tree location on the plot (i.e. distance from the plot centre and azimuth) were recorded.

 

FIGURE 3. Example of the sampling plot layout (yellow circles) within MODIS 1km pixels (red parallelograms). White lines mark borders of forest compartments (labels in italic) that are part of the management unit “Slavir”, part of Spačva Basin.

 

TABLE 3. The radius of sampling with respect to stand age and tree size.

 

Tree cores were taken, on average, from 9.6 dominant and co-dominant trees per plot (min. 5, max. 13). In total, 1051 cores were collected, out of which 383 in Pokupsko Basin (247 Q. robur, 21 C. betulus, 34 F. angustifolia, 75 A. glutinosa, 6 other) and 668 in Spačva Basin (512 Q. robur, 44 C. betulus, 112 F. angustifolia).

One core per tree was taken at 1.30 m from the ground from the stem side facing the plot centre using increment borer (Haglof, Sweden) of 5.15 mm inner diameter. The collected cores were air-dried in the laboratory for several days and stored in the refrigerator at 4°C until further analysis. Following standard dendrochronological preparation methods, outlined in Stokes and Smiley [41]), the cores were glued to wooden holders and placed into a press for a day. Afterward, they were sanded with progressively finer grades of sandpaper (i.e. 120, 180, 240 and 320 grit). Finally, cores were scanned at high resolution (2400 DPI) and the scanned images were saved into the tree-core database on a local network drive for later TRW measurements.

Tree-ring widths were measured from scanned images to the nearest 0.001 mm using PC and specialized CooRecorder software v.7.8.1 (Cybis Elektronik & Data AB, Sweden). TRW measurements were corrected and underwent quality control through repeated cross-check routines for cross-dating and identification of measurement errors with COFECHA computer program [42, 43].

Data Analysis

The comparison of growth dynamics from 2000 to 2014 in two distinct locations, from two different data sources, was performed using basal area growth estimated from tree-rings and net primary productivity obtained from the BBGCMuSo model. To exclude age-related trend associated with TRW data we used basal area increment (BAI) as proposed in Biondi and Qeadan [44]. BAI was calculated using tree diameter at breast height, measured at the time of tree coring, and TRW data.

Net primary productivity (NPP) obtained from the BBGCMuSo model comprises net productivities of different tree parts. To be able to make a comparison of BAI with the model NPP data, we estimated the net primary productivity of stem wood (NPPw) using carbon allocation ratios (Table 4) in the following way: 

NPP = NPPw + NPPl + NPPf + NPPfr + NPPcr

NPP = (1.42 + 1+ 0.14 + 0.95 + 0.26 · 1.42) · NPPl

NPP = 3.88 NPPl

NPPw = 1.42 NPPl

NPPw = 0.366 NPP

where NPP is net primary productivity, and subscripts w, l, f, fr and cr stand for wood, leaf, fruit, fine root and coarse root, respectively.

 

TABLE 4.Carbon allocation parameters and values used in BBGCMuSo model

 

Modelling results are area-based, i.e. NPPw is expressed in kg·C·m-2·y-1, while tree‑ring data are tree-based, i.e. BAI is expressed in cm2·y-1. To be able to assess the growth dynamics of the results from the modelling against the tree-ring data we calculated the average annual NPPw and the average annual BAI from all simulation runs (36 in Pokupsko basin and 55 in Spačva basin) and all tree cores for each of the forest areas, respectively. Then we standardized both NPPw and BAI. Standardization is introduced because a direct comparison of NPPw and BAI is not possible without introducing additional uncertainty. To make a comparison without the standardization we would need to calculate the net primary productivity of wood based on the tree core data. This would require the use of allometric functions for estimating wood volume, as well as the use of wood density and wood carbon content values. In addition, not all trees in the plots have been cored, and NPP of the uncored trees would have to be evaluated. All this would introduce additional errors. On the other hand, using standardized values, despite being somewhat more difficult to grasp, circumvents those problems and at the same time keeps the information on growth dynamics. 

Standardized values (z-values) of BAI or NPPw, were calculated as:

where  is the variable of interest (the average tree BAI or the average simulated NPPw) in the year  (i = 2000 to 2014) at the given area;  is the overall average of all tree BAI, or NPPw of all simulated forest compartments, at a given area during the entire 15-years long period of interest;  and   is the standard deviation of  during the period of interest (in our case years 2000 to 2014).

Before performing standardization, a Shapiro-Wilk W test for normality was conducted on a series of average annual tree BAI and average annual NPPw estimates for each forest area using procedure swilk in STATA 14 (StataCorp, College Station, TX, USA). Average annual BAI data series were normally distributed (W=0.9791, p=0.9630 for Pokupsko; W=0.9559, p=0.6224 for Spačva). Similarly, average NPPw data series were also normally distributed (W=0.9532, p=0.5756 for Pokupsko; W=0.9699, p=0.8563 for Spačva). Therefore, standardization of the data sets is allowed. 

Meteorological data were analysed for the same period as for growth dynamics, from 2000 to 2014. For each location mean annual (October-September) and seasonal (April‑September) air temperature (°C) and precipitation (mm) anomalies were calculated as follows:

T = Ti - Tp 

P = Pi - Pp

where T is air temperature anomaly, Ti is mean annual (Oct-Sep) or seasonal (Apr-Sep) air temperature in year i , Tp is mean annual/seasonal air temperature of the investigated period (2000-2014), P is precipitation anomaly, Pi is annual/ seasonal precipitation sum, and Pp is the mean of annual/ seasonal precipitation sums during the investigated period.

 

RESULTS AND DISCUSSION

Measured Growth Dynamics and Observed Meteorology

Tree-ring data revealed differences in growth dynamics between two investigated oak forests (Figure 4; green circles). Interestingly, at the wetter [19, 45, 46] oak site (Pokupsko), growth decreased in colder years (e.g. 2005-2006, Figure 5), which is in contrast to the common negative response of oak trees’ growth to temperature in spring and summer found in Čufar et al. [47], and to the positive response of oaks to rainy, humid and cloudy conditions during the current year’s summer [16]. Nevertheless, according to Renninger et al. [48], it is highly important to account for groundwater table information when interpreting the response of oak ecosystems to dry conditions. Due to low vertical water conductivity of gleysol soil, forests at Pokupsko site are partly flooded with stagnating water during winter and early spring. During a course of a vegetation season groundwater table is relatively high (Table 1), and therefore we can consider that at this particular site growth is rarely water‑limited, but can be rather sunlight-limited during colder cloudy years. For example, in Pokupsko basin in 2011 there was approx. 300 h more sun compared to 2010, or almost 17%. What is even more important, sun hours were in shortage during May and September of 2010 while the case of 2011 it was exactly the opposite (data for meteorological station Karlovac, http://klima.hr/klima_e.php?id=klima_elementi). Contrary to that, at the drier site (Spačva), growth decreased in warm and dry years (e.g. 2007), which is in line with Čufar et al. [47]. Forests at Spačva site can be considered to be water-limited during warm and dry years, especially after the prolonged drought from the ecological perspective, when groundwater table drops significantly (i.e. more than 1 m below the long-term average for a given month), although partial resupply of soil water reserves occurs laterally. In addition, forests in Spačva basin grow on fertile soil, with good water holding capacity, and are considered to be highly productive. Forests that grow on soil with high nutrient availability tend to have higher aboveground biomass and are found to be more susceptible to drought due to a predisposition to hydraulic failure [49]. At both sites model falsely indicated a drop in growth (z(NPPw)) for the dry 2011 (Figure 4). But in 2012, the growth decrease indicated by the model was evident also in tree core (z(BAI)). The observed reduction in growth was a consequence of two extremely warm and dry years in a row (i.e. 2011 and 2012) and is likely due to the carry-over effect of the drought [35, 50].

 

FIGURE 4. Measured and simulated growth dynamics during the period 2000-2014 in two distinct locations of pedunculate oak forests (Pokupsko and Spačva Basin) based on standardized values z(BAI) from tree cores, and z(NPPw) from BBGCMuSo model. The trends and the corresponding equations for measured and modelled z-values are also shown.

 

FIGURE 5. Mean annual (October from the previous year – September of the current year) and growing season (April-September from the current year) air temperature (°C) and precipitation (mm) anomalies during the period 2000-2014 in two distinct locations of pedunculate oak forests (Pokupsko and Spačva Basin).

 

Evaluating the Predicting Power of the Model

Model results show some differences in growth dynamics between two investigated locations (Figure 4; red triangles). A strong reduction in simulated growth in years 2003 and 2012, observed at both sites, indicates high model sensitivity to dry conditions and high air temperature, i.e. negative precipitation anomalies and positive temperature anomalies during vegetation period (Figure 5). Similar modelling results at both locations indicate that the model is more sensitive to meteorology than to site‑specific conditions. Although two locations have somewhat different abundances of main tree species, soil characteristics and hydrology, as well as the measured growth dynamics (Figure 4), modelled growth shows similar dynamics (Pearson’s correlation coefficient between NPPw (Pokupsko) and NPPw (Spačva) is 0.563). Pappas et al. [3] obtained similar results when testing process-based LPJ‑GUESS model. Authors concluded that model has a very high sensitivity to photosynthetic parameters (i.e. light correlated parameters) and minor sensitivity to hydrological and soil texture parameters.

Quantitative comparison of the growth dynamics from the two different data sources (the measured tree rings and BBGCMuSo model) reveals a poor agreement (i.e. correlation) for both sites (Figure 6). Table 5 shows the results of statistical evaluation for the model-measurement agreement. The extremely dry year 2003 acted as an outlier, according to Tukey’s definition [51], for NPPw at wetter (Pokupsko) sites (Figure 6, left panel). It seems that a single extremely dry year, such as the year 2003, when strongly negative effects on vegetation productivity at European scale were recorded [52], has not significantly affected tree growth at the investigated sites (Figure 4). The ability of oak trees to overcome a single dry event could be explained by the presence of significant soil water reserves (e. g. records from groundwater monitoring in Spačva show that depth to groundwater in Spačva Basin can fluctuate from 0 to ~5 m, while the average water holding capacity of soils is150 mm·m-1 [46]) and/or large carbohydrate reservoirs (i.e. carbon storage pools in trees). The analysis of remote sensing data also indicates a different response of forests and other vegetation to drought [35], where the results suggest that drought in a given year might negatively affect growth in the consecutive years in case of forests, but not for other vegetation types. This is in line with the logic that due to stress carbohydrate reserves might be depleted because of decreased photosynthesis (due to stomatal closure) and/or increased respiration demand due to excess heat, which then has a legacy effect on the growth in the next year [53].

 

TABLE 5. Performance statistics (based on z-values) of Biome-BGCMuSo compared with the observed tree ring data.

 

FIGURE 6. Assessing the predictive power of the model for two distinct locations of pedunculate oak forests at Pokupsko and Spačva Basin with standardized values (z-values) of BAI and NPPw.

 

Significantly decreased growth in 2003 obtained from the model indicates that model routines, describing a biological response to the single drought event, have difficulties with predicting growth dynamics under such conditions. In stress conditions carbon allocation ratios (i.e. proportions of assimilates allocated to different plant pools/organs, as well as mobilization of reserves) change in order for the plant to successfully overcome stress [54]. The shortcoming of fixed carbon allocation ratios, currently implemented in the BBGCMuSo model, might become increasingly pronounced in the case of extreme events. Additionally, BBGCMuSo is a “source-driven” model, meaning that the current photosynthates are immediately allocated to tissues with fixed allocation ratios. According to new findings (e.g. [54]), this logic might not be completely applicable to forests, which can partly explain why the extreme event in a given year might have pronounced effect on plant growth in the forthcoming year(s). These issues are limiting the model to properly predict plant’s response to drought stress. The improvement of modelling results for both sites might be achieved if groundwater table information is used [24].

Residual analysis was performed to find possible sources of model-data discrepancy. According to Figures 7 and 8, the relationship between the studied meteorological variables (data from the previous year and the current year) and the model residuals was not significant for the two study sites. However, the trends, although not statistically significant, might be indicative. Positive/negative precipitation (Figure 7) anomaly in the current year seems to cause over/underestimation of NPPw at both sites, although data variability is high. This means that the role of water availability is more emphasized in the model than in reality. On the other hand, there is a difference between locations in the model performance for the current year with respect to the precipitation anomaly in the previous year (Figure 7). At the wetter (Pokupsko) location negative precipitation anomaly still seems to cause the underestimation of the NPPw by the model, but at the drier site (Spačva) the effect is reversed (negative precipitation anomaly seems to cause overestimation of NPPw by the model). Opposite to the effects of precipitation, positive/negative air temperature (Figure 8) anomaly in the current year seems to have different effects at each location. At the wetter (Pokupsko) location, positive temperature anomaly seems to cause underestimation of NPPw, while at the drier (Spačva) site it causes overestimation. Interestingly, positive air temperature anomaly in the previous year seems to cause underestimation of NPPw at both locations. The observed relations are not statistically significant, as we already emphasized, but are in accordance with the logic of buffered growth response against single drought events due to a large rooting depth of trees.

 

FIGURE 7. The correlation of residuals with seasonal (April-September) precipitation anomaly during the previous and the current year for Pokupsko (a) and Spačva Basin (b).

 

FIGURE 8. The correlation of residuals with seasonal (April-September) temperature anomaly during the previous and the current year for Pokupsko (a) and Spačva Basin (b).

 

In a previous study [24] statistical evaluation of the observed Biome-BGCMuSo simulated carbon fluxes against measured eddy covariance data from Pokupsko Basin (Jastrebarsko site) showed much better agreement (see Table 5 in [24]) than in the current study. The explained variance of the observed gross primary production (GPP) and total ecosystem respiration (TER) reached 84% and 83%, respectively. This is in striking contrast to the negligible amount of explained variance (~2%) for the BAI dataset. It is important to note that biogeochemical models like Biome-BGCMuSo are typically calibrated/validated with data-rich eddy covariance based measurements. The application of BAI as validation data in the case of processed based models is rare in the literature. The use of BAI data or NPP estimated from BAI measurements, in Biome-BGCMuSo calibration might help to improve the allocation parameters and thus improve the predictive power of the model. This would be even more important when dynamic allocation will be implemented in the next version of Biome-BGCMuSo (Barcza, personal communication). Multi-objective calibration using both eddy covariance and BAI data (probably with additional variables such as leaf mass, leaf C:N ratio) might provide additional constraints to improve model performance in such cases. The calibration will be a challenging task where sophisticated calibration (e.g. Bayesian [55]) techniques will have to be used. 

 

CONCLUSIONS

The comparison of modelling results with the observed tree-ring data revealed two important model issues related to its predictive power. The first one is the importance of including site-specific conditions (i.e. groundwater table information) with the purpose of enabling the model to be more case-sensitive. At both oak forest locations, Pokupsko and Spačva basins, BBGCMuSo showed poor predictive power in capturing the dynamics obtained from tree‑ring data. Using groundwater table information for modelling in lowland oak forests might improve model results.

The second issue is related to carbon allocation. Fixed carbon allocation ratios, currently used in BBGCMuSo model, do not enable the model to successfully predict plant response to stress conditions (e.g. drought). A dynamic carbon allocation routine might better capture tree stress response and growths dynamics. There is an urgent need to investigate and implement more sophisticated carbon allocation routines in the BBGCMuSo model.

 

Acknowledgments

The research has been supported by the Croatian Science Foundation (HRZZ UIP-11-2013-2492) and “Spačva” Project (OKFŠ, HŠ 2013-2016). The work of Z.B. and D.H. on the research was funded by the Széchenyi 2020 programme, the European Regional Development Fund and the Hungarian Government (GINOP-2.3.2-15-2016-00028) and Z.B. was also supported by the grant “EVA4.0”, No. CZ.02.1.01/0.0/0.0/16_019/0000803 financed by OP RDE.

We would like to thank two anonymous reviewers for their comments which greatly improved the manuscript.

 


REFERENCES

  1. SÁNCHEZ-SALGUERO R, CAMARERO JJ, GUTIÉRRET E, GONZÁLEZ RF, GAZOL A, SANGÜESA-BARREDA G, ANDREU-HAYLES L, LINARES JC, et al. 2017 Assessing forest vulnerability to climate warming using a process-based model of tree growth: bad prospects for rear-edges. Glob Change Biol 23 (7): 2705-2719. DOI: https://doi.org/10.1111/gcb.13541
  2. PETERS EB, WYTHERS KR, ZHANG S, BRADFORD JB, REICH PB 2013 Potential climate change impacts on temperate forest ecosystem processes. Can J For Res 43 (10): 939-950. DOI: https://doi.org/10.1139/cjfr-2013-0013
  3. PAPPAS C, FATICHI S, LEUZINGER S, WOLF A, BURLANDO P 2013 Sensitivity analysis of a process-based ecosystem model: Pinpointing parameterization and structural issues. J Geophys Res Biogeosci 118 (2): 505-528. DOI: https://doi.org/10.1002/jgrg.20035
  4. FATICHI S, LEUZINGER S, KÖRNER C 2014 Moving beyond photosynthesis: from carbon source to sink-driven vegetation modelling. New Phytol 201 (4): 1086-1095. DOI: https:// doi.org/10.1111/nph.12614
  5. HARTIG F, DYKE J, HICKLER T, HIGGINS SI, O’HARA RB, SCHEITER S, HUTH A 2012 Connecting dynamic vegetation models to data – an inverse perspective. J Biogeogr 39 (12): 2240-2252. DOI: https:// doi.org/10.1111/j.1365-2699.2012.02745.x
  6. BALDOCCHI DD 2003 Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: past, present and future. Glob Change Biol 9 (4): 479-492. DOI: https:// doi.org/10.1046/j.1365-2486.2003.00629.x
  7. BALDOCCHI DD, FALGE E, GU L, OLSON R, HOLLINGER D, RUNNING SW, ANTHONI P, BERNHOFER C, et al. 2001 FLUXNET: A New Tool to Study the Temporal and Spatial Variability of Ecosystem-Scale Carbon Dioxide, Water Vapor, and Energy Flux Densities. B Am Meteorol Soc 82 (11): 2415-2434. DOI: https://doi.org/10.1175/1520-0477(2001)082<2415:FANTTS>2.3.CO;2
  8. KUMAR J, HOFFMAN FM, HARGROVE WW, COLLIE RN 2016 Understanding the representativeness of FLUXNET for upscaling carbon flux from eddy covariance measurements. Earth Syst Sci Data Discuss. Manuscript under review. DOI: https://doi.org/10.5194/essd-2016-36
  9. BARCZA Z, KERN A, HASZPRA L, KLJUN N 2009 Spatial representativeness of tall tower eddy covariance measurements using remote sensing and footprint analysis. Agr Forest Meteorol 149 (5): 795-807. DOI: https://doi.org/10.1016/j.agrformet.2008.10.021
  10. EVANS MEK, FALK DA, ARIZPE A, SWETNAM TL, BABST F, HOLSINGER DKE 2017 Fusing tree-ring and forest inventory data to infer influences on tree growth. Ecosphere 8 (7): e01889. DOI: https://doi.org/10.1002/ecs2.1889
  11. BÜNTGEN U, FRANK DC, WILSON R, CARRER M, URBINATI C, ESPER J 2008 Testing for tree-ring divergence in the European Alps. Glob Change Biol 14 (10): 2443-2453. DOI: https://doi.org/10.1111/j.1365-2486.2008.01640.x
  12. ESPER J, FRANK DC, BÜNTGEN U, VERSTEGE A, LUTERBACHER J, XOPLAKI E 2007 Long-term drought severity variations in Morocco. Geophys Res Lett 34 (17): . DOI: https://doi.org/10.1029/2007GL030844
  13. FRITTS HC 1976 Tree Rings and Climate. Academic Press, London, UK, 567 p
  14. CARRER M, URBINATI M 2006 Long-term change in the sensitivity of tree-ring growth to climate forcing in Larix decidua. New Phytol 170 (4): 861-872. DOI: https://doi.org/10.1111/j.1469-8137.2006.01703.x
  15. SCHARNWEBER T, MANTHEY M, CRIEGEE C, BAUWE A, SCHRÖDER C, WILMKING M 2011 Drought matters - declining precipitation influences growth of Fagus sylvatica and Quercus robur L. in north-eastern Germany. Forest Ecol Manag 262 (6): 947-961. DOI: https://doi.org/10.1016/j.foreco.2011.05.026
  16. HAFNER P, GRIČAR J, SKUDNIK M, LEVANIČ T 2015 Variations in Environmental Signals in Tree-Ring Indices in Trees with Different Growth Potential. PLoS ONE 10 (11): e0143918. DOI: https://doi.org/10.1371/journal.pone.0143918
  17. STOJANOVIĆ DB, LEVANIČ T, MATOVIĆ B, ORLOVIĆ S 2015 Growth decrease and mortality of oak floodplain forests as a response to change of water regime and climate. Eur J For Res 134 (3): 555-567. DOI: https://doi.org/10.1007/s10342-015-0871-5
  18. WRB 2006 World reference base for soil resources 2006. A framework for international classification, correlation and communication. World Soil Resources Reports No. 103, FAO, Rome, Italy, 128 p
  19. MAYER B 1996 Hydropedological relations in the region of lowland forests of the Pokupsko basin (in Croatian with English summary). Rad Šumar Inst 31 (1-2): 37-89
  20. SELETKOVIĆ Z 1996 Climate in Pedunculate oak forests (in Croatian with English summary). In: Klepac D (ed) Hrast lužnjak u Hrvatskoj. HAZU - Centar za znanstveni rad Vinkovci i "Hrvatske šume" d.o.o., Vinkovci-Zagreb, Croatia, pp 71-82
  21. KALINIĆ M 1975 Soils of forest communities in Spačva basin (in Croatian with English summary). In: Varićak T, Vidaković M, Švagelj D (eds) Sto godina znanstvenog i organiziranog pristupa šumarstvu jugoistočne Slavonije. HAZU Centar za znanstveni rad Vinkovci, Zagreb, Croatia, pp 413-432
  22. PILAŠ I 2017 Climatic, hydro-pedologic and geomorphologic characteristics of Spačva basin. In: Final report of the project "Ecological-climate changes and regeneration challenges of Pedunculate oak forests in Spačva basin" (in Croatian). Croatian Forest Research Institute, Jastrebarsko, Croatia, pp 161-195
  23. HIDY D, BARCZA Z, HASZPRA L, CHURKINA G, PINTÉR K, NAGY Z 2012 Development of the Biome-BGC model for simulation of managed herbaceous ecosystems. Ecol Model 226: 99-119. DOI: https://doi.org/10.1016/j.ecolmodel.2011.11.008
  24. HIDY D, BARCZA Z, MARJANOVIĆ H, OSTROGOVIĆ SEVER MZ, DOBOR L, GELYBÓ G, FODOR N, PINTÉR K, et al. 2016 Terrestrial Ecosystem Process Model Biome-BGCMuSo v4.0: Summary of improvements and new modelling possibilities. Geosci Model Dev 9 (12): 4405-4437. DOI: https://doi.org/10.5194/gmd-9-4405-2016
  25. THORNTON PE 2000 User’s Guide for Biome-BGC, Version 4.1.1. URL: ftp://daac.ornl.gov/data/model_archive/BIOME_BGC/biome_bgc_4.1.1/comp/bgc_users_guide_411.pdf
  26. PIETSCH SA, HASENAUER H, KUCERA J, CERMÁK J 2003 Modeling effects of hydrological changes on the carbon and nitrogen balance of oak in floodplains. Tree Physiol 23 (11): 735-746
  27. CIENCIALA E, TATARINOV FA 2006 Application of BIOME-BGC model to managed forests. 2. Comparison with long-term observations of stand production for major tree species. Forest Ecol Manag 237 (1-3): 252-266. DOI: https://doi.org/10.1016/j.foreco.2006.09.086
  28. HLÁSNY T, BARCZA Z, FABRIKA M, BALÁZS B, CHURKINA G, PAJTÍK J, SEDMÁK R, TURCÁNI M 2011 Climate change impacts on growth and carbon balance of forests in Central Europe. Climate Res 47 (3): 219-236. DOI: https://doi.org/10.3354/cr01024
  29. MERGANIČOVÁ K, MERGANIČ J 2014 The Effect of Dynamic Mortality Incorporated in BIOME-BGC on Modelling the Development of Natural Forests. J Environ Inform 24 (1): 24-31. DOI: https://doi.org/10.3808/jei.201400273
  30. HLASNY T, BARCZA Z, BARKA I, MERGANIČOVÁ K, SEDMAK R, KERN A, PAJTIK J, BALAZS B, FABRIKA M, CHURKINA G 2014 Future carbon cycle in mountain spruce forests of Central Europe: Modelling framework and ecological inferences. Forest Ecol Manag 328: 55-68. DOI: https://doi.org/10.1016/j.foreco.2014.04.038
  31. DOBOR L, BARCZA Z, HLÁSNY T, HAVASI Á, HORVÁTH F, ITTZÉS P, BARTHOLY J 2015 Bridging the gap between climate models and impact studies: The FORESEE Database. Geosci Data J 2 (1): 1-11. DOI: https://doi.org/10.1002/gdj3.22
  32. BINKLEY D, CROMAC Kjr, BAKER DD 1994 Nitrogen Fixation by Red Alder: Biology, Rates, and Controls. In: Hibbs DE, DeBell DS, Tarrant RF (eds) The Biology and Management of Red Alder. Oregon State University Press, Corvallis. Or (xi) 256 p
  33. MARJANOVIĆ H, OSTROGOVIĆ MZ, ALBERTI G, BALENOVIĆ I, PALADINIĆ E, INDIR K, PERESOTTI A, VULETIĆ D 2011 Carbon dynamics in younger stands of pedunculate oak during two vegetation periods (in Croatian with English summary). Sumar list 135 (11-12): 59-73
  34. Rulebook on forest management 2015 Official Gazette of the Republic of Croatia 79/15.
  35. KERN A, MARJANOVIĆ H, DOBOR L, ANIĆ M, HLÁSNY T, BARCZA Z 2017 Identification of Years with Extreme Vegetation State in Central Europe Based on Remote Sensing and Meteorological Data. South-east Eur for 8 (1): 1-20. DOI: https://doi.org/10.15177/seefor.17-05
  36. MIKO S, HASAN O, KOMESAROVIĆ B, ILJANIĆ N, ŠPARICA MIKO M, ĐUMBIR AM, OSTROGOVIĆ SEVER MZ, PALADINIĆ E, MARJANOVIĆ H 2017 Promjena zaliha ugljika u tlu i izračun trendova ukupnog dušika i organskog ugljika u tlu te odnosa C:N (Soil carbon stock change and calculation of trends of total soil nitrogen, organic carbon, and C:N ratio - in Croatian), Study for the Croatian Agency for the Environment and Nature (Contract No. 10-14-1442/79). Zagreb: Croatian Geological Survey, Jastrebarsko: Croatian Forest Research Institute, Zagreb: Agricultural Land Agency, 81p
  37. ETHERIDGE DM, STEELE LP, LANGENFELDS RL, FRANCEY RJ, BARNOLA JM, MORGAN VI 1996 Natural and anthropogenic changes in atmospheric CO2 over the last 1000 years from air in Antarctic ice and firn. J Geophys Res 101 (D2): 4115-4128. DOI: https://doi.org/10.1029/95JD03410
  38. CHURKINA G, BROVKIN V, VON BLOH W, TRUSILOVA K, JUNG M, DENTENER F 2009 Synergy of rising nitrogen depositions and atmospheric CO2 on land carbon uptake moderately offsets global warming. Global Biogeochem Cy 23 (4): GB4027. DOI: https://doi.org/10.1029/2008GB003291
  39. RUNNING SW, ZHAO M 2015 User’s Guide Daily GPP and Annual NPP (MOD17A2/A3) Products NASA Earth Observing System MODIS Land Algorithm, Ver. 3.0 for Coll. 6, 7 October 2015. URL: http://www.ntsg.umt.edu/files/modis/MOD17UsersGuide2015_v3.pdf (15 Nov 2017).
  40. OSTROGOVIĆ MZ 2013 Carbon stocks and carbon balance of an even-aged Pedunculate Oak (Quercus robur) forest in Kupa river basin. Ph.D. thesis, University of Zagreb, Faculty of Forestry, 130 pp. (in Croatian with English summary). URL: https://dr.nsk.hr/en/islandora/object/sumfak%3A795 (Nov 2017).
  41. STOKES MA, SMILEY TL 1968 An Introduction to tree-ring dating. Chicago: University of Chicago Press, USA, 73 p
  42. HOLMES RL 1983 Computer-assisted quality control in tree-ring dating and measurements. Tree-Ring Bull 43: 69-78
  43. GRISSINO-MAYER HD 2001 Evaluating crossdating accuracy: a manual and tutorial for the computer program COFECHA. Tree-Ring Res 57: 205-221
  44. BIONDI F, QEADAN F 2008 A theory-driven approach to tree-ring standardization: defining the biological trend from expected basal area increment. Tree-Ring Res 64 (2): 81-96. DOI: https://doi.org/10.3959/2008-6.1
  45. PERNAR N, BAKŠIĆ D 2005 The soil of floodplain forests. In VUKELIĆ J (ed) Floodplain forests in Croatia, Akademija šumarskih znanosti & Hrvatske šume & Grad Zagreb, Zagreb, Croatia, pp 71-85.
  46. VRBEK B, PILAŠ I, PERNAR N 2011 Observed climate change in Croatia and its impact on the hydrology of lowlands. In: Bredemeier M, Cohen S, Godbold Dl, et al. (eds) Forest management and the water cycle: An ecosystem-based approach, Springer Netherlands, Dordrecht, Netherlands, pp 141-162. DOI: https://doi.org/10.1007/978-90-481-9834-4_8
  47. ČUFAR K, GRABNER M, MORGÓS A, CASTILLO EM, MERELA M, DE LUIS M 2014 Common climatic signals affecting oak tree-ring growth in SE Central Europe. Trees 28 (5): 1267-1277. DOI: https://doi.org/10.1007/s00468-013-0972-z
  48. RENNINGER HJ, CARLO N, CLARK KL, SCHÄFER KVR 2014 Physiological strategies of co-occurring oaks in a water- and nutrient-limited ecosystem. Tree Physiol 34 (2): 159-173. DOI: https://doi.org/10.1093/treephys/tpt122
  49. GESSLER A, SCHAUB M, MCDOWELL NG 2017 The role of nutrients in drought-induced tree mortality and recovery. New Phytol 214 (2): 513-520. DOI: https://doi.org/10.1111/nph.14340  
  50. MOLEN MK, DOLMAN AJ, CIAIS P, EGLIN T, GOBRON N, LAW BE, MEIR P, PETERS W, et al. 2011 Drought and ecosystem carbon cycling. Agr Forest Meteorol 151 (7): 765-773. DOI: https://doi.org/10.1016/j.agrformet.2011.01.018
  51. TUKEY JW 1977 Exploratory Data Analysis. Addison-Wesley. ISBN-13: 978-0201076165, 688 p
  52. CIAIS PH, REICHSTEIN M, NICOLAS V, GRANIER A, OGÉE J, ALLARD V, AUBINET M, BUCHMANN N, et al. 2005 Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature 437: 529-533. DOI: https://doi.org/10.1038/nature03972
  53. ZHAO J, HARTMANN H, TRUMBORE S, ZIEGLER W, ZHANG Y 2013 High temperature causes negative whole-plant carbon balance under mild drought. New Phytologist 200 (2): 330-339. DOI: https://doi.org/10.1111/nph.12400
  54. HAGEDORN F, JOBIN J, PETER M, LUSTER J, PRITSCH K, NICKEL UT, KERNER R, MOLINIER V, et al. 2016 Recovery of trees from drought depends on belowground sink control. Nat Plants 2: 16111. DOI: https://doi.org/10.1038/nplants.2016.111
  55. VAN OIJEN M, ROUGIER J, SMITH R 2005 Bayesian calibration of process-based forest models: Bridging the gap between models and data. Tree Physiology 25 (7): 915-927. DOI: https://doi.org/10.1093/treephys/25.7.915
  56. MORIASI DN, ARNOLD JG, VAN LIEW MW, BINGNER RL, HARMEL RD, VEITH TL 2007 Model Evaluation Guidelines for Systematic Quantification of Accuracy in Watershed Simulations. T ASABE 50 (3): 885–900. DOI: https://doi.org/10.13031/2013.23153

© 2017 by the Croatian Forest Research Institute. This is an Open Access paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0).

 

TABLE 5. Performance statistics (bas

- Download PDF-

SEEFOR 8 (2): 127-136
Article ID: 173
DOI: https://doi.org/10.15177/seefor.17-18

Original scientific paper

 

Structure, Yield and Acorn Production of Oak (Quercus robur L.) Dominated Floodplain Forests in the Czech Republic and Croatia


Lumír Dobrovolný1*, Antonín Martiník2, Damir Drvodelić3, Milan Oršanić3


(1) Mendel University in Brno, School Forest Enterprise Křtiny, Zemědělská 3, CZ-61300 Brno, Czech Republic;
(2) Mendel University in Brno, Faculty of Forestry and Wood Technology, Department of Silviculture, Zemědělská 3, CZ-61300 Brno, Czech Republic;
(3) University of Zagreb, Faculty of Forestry, Department of Forest Ecology and Silviculture, Svetošimunska 25, HR-10000 Zagreb, Croatia

* Correspondence: e-mail:

Citation: DOBROVOLNÝ L, MARTINÍK A, DRVODELIĆ D, ORŠANIĆ M 2017 Structure, Yield and Acorn Production of Oak (Quercus robur L.) Dominated Floodplain Forests in the Czech Republic and Croatia. South-east Eur for 8 (2): 127-136. DOI: https://doi.org/10.15177/seefor.17-18

Received: 9 Nov 2017; Revised: 7 Dec 2017; Accepted: 14 Dec 2017; Published online: 21 Dec 2017


Cited by:     Crossref      Google Scholar


Abstract

Background and Purpose: The study aims at comparing two (over 100 years old) floodplain oak-predominated forests in the Czech Republic (CZ) with two in of Croatia (HR) with regards to: i) their structure and yield and, more specifically, ii) individual oak tree characteristics including acorn production.
Materials and Methods: In both countries a different silvicultural concept is preferred (CZ: clear-cutting management with artificial regeneration, HR: shelterwood management with natural regeneration). The main research goal was to create a basic decision tool for forest managers and open some questions for future research.
Results: Despite the different natural and management practices, the total standing volume of floodplain forest was found to be similar in both countries, ranging from 500 to 700 m3·ha-1 (basal area: 34-41 m2·ha-1). In CZ generally more poor structure diversity was detected. Although in CZ the number of crop oaks (130-160 oaks per hectare) was almost double as compared with HR, the CZ oaks had shorter crowns, almost twice smaller crown projection, lower mean volume and lower share of valuable assortments.
Conclusions: Despite the total standing volume of oaks in HR being lower than in CZ, the total yield was observed in Croatia (loss in CZ ca. 22,000 €·ha-1). The acorn density and quality were generally higher in HR with a more even distribution as well. Despite more favourable climatic conditions in HR, the currently used management system in CZ floodplain forests should be gradually converted to the Croatian model with a multi-layered forest structure, more focused on individual tree growth and stability with high economical value and high reproductive potential.

Keywords: floodplain forest, silvicultural system, pedunculate oak, structure diversity, assortment structure of oak, yield, acorn production



INTRODUCTION

Pedunculate oak (Quercus robur L.) is considered to be one of the most important economic tree species in floodplain forests in Europe [1]. Natural regeneration of oak, as in the case of other tree species, is a complex process influenced by many biotic and abiotic factors. The main negative factors are fungal infections and diseases (Microsphaera alphitoides), consumption by animals (insects, birds, rodents and wild boars), light, water availability and climatic factors such as late frosts [2]. From a different viewpoint, the social position and individual growth characteristics of oak trees such as growing space, crown size and architecture are included amongst the key factors for the abundance and quality of acorns [2-5] and for highly valuable timber production as well [6]. In this context, the forest structure (species and spatial diversity) and its targeted management can significantly influence the successfulness ofnatural forests.

In natural forests, pedunculate oak with its high-age growth strategy has enough time and space to create great stem and crown dimensions. Here, one oak generation equals two to four generations of hornbeam and other accompanying species. This makes the spatial structure of natural floodplain forests relatively rich and dynamic in time [7-8].

A predominant silvicultural system in floodplain forests in the Czech Republic (CZ) is clear-felling (with a maximum size of 2 ha) with mechanical soil preparation and artificial regeneration [9]. Reasons for this are: insufficient acorn crops, strong weed competition and high impact of small and big vertebrates [10]. This management results in the floodplain forest structure being less diverse and more homogeneous with a high number of trees in the overstorey and under-developed crowns with poor fructification in adult age [3]. On the other hand, in a number of cases in the southern part of CZ, Dobrovolný [11] and Martiník et al. [3] demonstrated success of natural regeneration of oak if certain conditions were met.

Matić [12] and Oršanić and Drvodelić [13] consider pedunculate oak to be a tree species with a climax strategy and recommend, the traditionally used natural regeneration of oak under the shelter of the mother stand in three or two cuts. This method takes into account the biological and economic properties and ecological requirements of acorns and causes minimal stress to the soil and the stand [1, 14]. Diaci et al. [27] admit even irregular group felling in floodplain forests in Slovenia.

This study was focused on comparing two types of management of adult floodplain oak-predominated forests in the Czech Republic and Croatia with regards to: i) their structure and yield and, more specifically, ii) oak individual tree characteristics and acorn production. The main research goal was to create a basic decision tool for forest managers and to open some questions for future research.

 

MATERIAL AND METHODS

In each country (Czech Republic “CZ”, Republic of Croatia “HR”) in the year 2013, two managed adult floodplain forest stands before regeneration felling that represented a typical species and spatial structure of that experimetal region were selected (Table 1). The selected stands in CZ and HR differed primarily in species composition (CZI - oak and ash, CZII - oak, HRI - oak and ash, HRII - oak, ash and hornbeam).

 

{modal images/arhiva/vol8_no2/drvodelic/t1.jpg|title=} TABLE 1. Basic information on research stands.

 

In CZ, specifically in the South Moravian region (Židlochovice), the research was conducted in floodplain forests (managed by the State), located along the Morava, Dyje, Svratka and Jihlava rivers. The predominating soil type, cambic fluvisol, was slightly gleyic, eubasic in double substrates with chernozem fossils (ca. from 160 cm) on fluvial Holocene sediments. In HR, the research was conducted in the floodplain forests of pedunculate oak within the area of the Sava River. Research encompassed the management unit of “Opeke” (managed by the Faculty of Forestry of the University of Zagreb). The dominant soils included pseudogley level terrains, deep, illimerised brown soil, pseudogleyic, and eugley epigleyic (in micro depressions). A comparison of long-term and short-term HR and CZ climatic data (Table 1, Figure 1) indicates a higher average annual (and monthly) temperature and annual (and monthly) amount of rainfall in HR.

 

FIGURE 1. Development of the average monthly temperature and average monthly amount of rainfall in CZ and HR (years 2000 - 2012). A comparison of climatic data indicates a higher average annual (and monthly) temperature and annual (and monthly) amount of rainfall in HR.

 

In each of the four selected stands, one representative circular research plot (RP) of 0.25 ha in size was established - CZ I, CZ II, HR I, HR II (Tab. 1). Within each RP, the following variables were collected for all trees with a diameter at breast height (DBH) of more than 7 cm: the coordinates (using Field-Map technology-Institute of Forest Ecosystem Research Ltd., Czech Republic), DBH, tree height (h), crown length (CL) - difference between the tree height and crown base and crown projection (CP). Outside the RP, all the crop oaks (with a DBH above 30 cm) that extended with their crowns into the RP and could affect the abundance of fallen acorns were also measured.

Model height curve was constructed according to the Michajlov formula [15]. The stem volume without bark (V) was calculated using volume functions [16]. Canopy cover (CC) of the stand was calculated as sum of individual crown projections. The competitive situation or space surrounding each of crop oaks was evaluated as the mean distance (D) of the targeted crop oak to all nearest neighbours of any species (with tree height over 20 m). The following indices of forest structure were calculated with the aid of BWINPro 6.3 (Northwest German Forest Research Station, Germany) [17]: (1) indices of species diversity: (1.a) Shannon index (SI) and (1.b) Evenness (EI) (standardized Shannon) based on the abundance of the species (depending on the number of trees (N) and the basal area (G) - the higher the values, the greater the diversity, (1.c) Species-profile index (API) based on species abundance in three height stand layers - the higher the values, the greater the diversity; (2) indices based on the spatial pattern ofthe zero tree and its three nearest neighbours: (2.a) Mixing index (MI) - the values express the spatial species diversity of each situation (MI=0.00 - all trees belong to the same species, MI=0.33 - one tree belongs to a different species, MI=0.67 - two trees belong to a different species, MI=1.00 - all trees belong to a different species), (2.b) Index of DBH differentiation (DI) (0.0-0.3 = no or low differentiation, 0.3-0.5 = medium differentiation, 0.5-0.7 = high differentiation, 0.7-1 = very high differentiation), (2.c.) Index of DBH dominance (DDI) (the higher the positive values, the greater the dominance of the zero tree over its neighbours; values nearing 0 indicate an indifferent relation and the higher the negative values, the greater the suppression of the zero tree). The spatial patternof crop oaks with the distance to the nearest neighbouring oak was evaluated in ArcGis 10 (Esri, Inc., USA) according to the formula of Clark and Evans [25].

The assortment structure of oak was assessed according to assortment tables of Dejmal [26]. The stems were sorted according to class: I. sliced venner log; II. peeled lower quality veneer log; III.a and III.b saw log; V. pulpwood; VI. fuelwood. The economic profit was assessed according to volume and the current Czech price list of oak assortments.

Acorn abundance was evaluated using 36 seed traps (round wire hoops with collection sacks) per RP, 0.25 m2 each (r= 0.28 m), arranged in a lattice format and placed 0.5 m above the ground. The spatial coordinates of all seed traps were measured. All the seed traps were installed before the acorns began to fall in September 2013. They were collected every two weeks. The amount of collected acorns was evaluated as the amount of seeds per square metre and the germination capacity was tested according to the Czech and ISTA standards [18].

The statistical differences between the RPs in terms of tree or acorn characteristics were tested using the non-parametric Kruskal-Wallis one-way analysis of variance (using Statistica 10 - StatSoft, Inc., USA). The spatial pattern of acorns (see Figure 7) in 5 categories of density (0; 0.1-5; 5.1-15; 15.1-30; 30.1-50 acorns per m2) was estimated on each RP using ArcGis 10 with Kernel statistics (interpolation) tool (Esri, Inc., USA). This analysis was a basis for deriving the share of area covered by acorns and the share of crop oaks belonging to different categories of acorn density (see Table 10).

 

RESULTS

Structure and Yield

The standing volume and the basal area were found to be similar in both countries. Surprisingly, for two stands (RP CZ I and HR I) both values were identical - about 700 m3·ha-1 or 40 m2·ha-1 (Table 2). As expected, CZ II (with its poor structure) had the lowest total number of trees and the smallest basal area and standing volume.

 

TABLE 2. Basic inventory data of research plots.

 

The highest number of trees per hectare on CZ I was due to the high number of trees (namely ash) in the lowest diameter classes (Figure 2). The total canopy cover was always higher in HR (over 100%), mainly due to the presence of the middle tree layer (Figure 2 and 4).With respect to tree species composition (Table 3), in CZ, it was ash that dominated on CZ I and oak on CZ II in terms of the number of trees, while in HR it was alder with oak that dominated on HR I and hornbeam on HR II. In terms of basal area and standing volume it was oak that dominated in both countries; however, the total standing volume of oak was higher in CZ.

 

FIGURE 2. Distribution of DBH classes of all tree species. The diameter distribution in HR was more broad compared to CZ.

 

FIGURE 3. Distribution of DBH classes of oak. In HR, oaks were relatively evenly represented within a wider range of diameters, while in CZ oaks are clustered into several diameter classes around 50 cm.

 

FIGURE 4.Distribution of height classes of all species. The height structure was more diverse in HR with at least three distinct tree layers.

 

TABLE 3. The share of species composition in regard to the number of trees (N), basal area (BA) and volume (V).

 

The most diverse species structure in HR was found for HR I (a total of six species), while in the CZ hornbeam and alder were absent in all cases. A more diverse species structure (even the vertical profile) in HR was confirmed also by the structure indices (SI, EI, API, MI) (Table 4).

 

TABLE 4.Structure indices.

 

The diameter distribution of all species in HR was broader compared to CZ (Figure 2). While in CZ the single- or double-peak distribution indicates the highest representation of trees of moderate thickness or very thin (CZ I) trees (Figure 2), in HR single-peak distribution with the highest representation of thin trees was observed. In HR, oaks were relatively evenly represented within a wider range of diameters, while in CZ oaks were clustered into several diameter classes around 50 cm (Figure 3). Nevertheless, the values of the diameter DI (Table 4) show a relatively high spatial variability on all RPs, unlike the homogenous CZ II. The DDI index indicates a more neutral relationship among the trees.

Similarly, the height structure was more diverse in HR with at least three distinct tree layers formed, the heights being about 10 m, 24 m, and 36 m (Figure 4). In CZ I, only two layers were formed (around 14 m and 36 m) and in CZ II only a single layer was formed (ca. 30 m).

Differences between the countries in oak assortments and values are shown in Table 5. The greatest differences between countries were in the share of valuable assortments (classes I and II). There is only about 5% of this class in CZ and about 20-30% in HR. Despite the total volume of oak in HR being lower than in CZ, the total yield was higher (loss in CZ - ca. 22,000 €·ha-1).

 

TABLE 5. Share of oak assortments and assessment of economical value.

 

Crop Oaks (DBH > 30 cm)

HR was found to contain less crop oaks compared with CZ (Table 6). In general crop oaks in HR reached higher mean DBH and V (Table 6); however, significantly only when CZ I was compared with HR II and CZ II with HR I and HR II (Table 9). Tree heights were similar, except for CZ II. Crop oak crown characteristics, i.e. CL and CP, were significantly greater in HR (except for CL of CZI when compared with HRI); the mean CP here being almost twice as large as those in CZ (Table 6 and 9). Mean distance (D) of the targeted crop oak to the nearest neighbours was greater in HR (significantly only when CZ II was compared with HR I and II) (Table 6 and 9). These results confirmed also the growth relationships (DBH vs. h vs. CP) of CZ-HR oak trees with the similar direction and shape of constructed curves (Figure 5 and 6).

 

TABLE 6. Individual crop oak characteristics (mean, SD - standard deviation, min - minimum, max - maximum values).

 

FIGURE 5. Height / diameter curves of CZ-HR crop oaks. HR oaks showed better height growth, but with the similar shape of both curves.

 

FIGURE 6. Relationships between DBH and crown projection "CP" of CZ-HR crop oaks. HR oaks showed larger crowns, but with the similar direction of both curves.

 

In both countries, oaks showed the same significant and even distribution across the plot with spacing (i.e. the distance to the nearest neighbouring oak) being greater for HR (Table 7).

 

TABLE 7. Distribution pattern of crop oaks.

 

Acorn Production

In CZ I, acorn density was the lowest of all plots (Table 8), but not significantly when compared with CZ II (Table 9). In CZ II, the density was significantly lower only when compared with HR I. Differences in acorn abundance in HR were not statistically significant. Despite very different germination rates in each stand, there were generally more germinable seeds per square metre in HR. In HR, more even spatial coverage of the totalarea by acorns was found compared with CZ (Figure7).

 

TABLE 8. Acorn characteristics (mean, SD - standard deviation, min - minimum, max - maximum values).

 

TABLE 9. The results of Kruskal-Wallis analysis of variance.

 

FIGURE 7. Spatial density of acorns with in the RPs. In HR, more even spatial cover age of area by acorns with heigher density was found compared with CZ.

 

The share of covered area by acorns was higher in HR (about 80% of the total area covered by medium or high density of acorns) compared with CZ (about 50-80% of the total area covered by null or low density of acorns) (Table 10). In HR medium or high acorn density / seedfall was observed in 80% of all oak trees, while in CZ null or weak seedfall was observed in 50-80% of all oaks (Table10).

 

TABLE 10.The share of area (1 RP = 0.25 ha) covered by different categories of acorn density and the share of crop oaks belonging to different categories of acorn density.

 

DISCUSSION

Our results based on comparisons of forest structure, yield and acorn production in HR - CZ cleared the way for basic silvicultural decisions (Table 11).

 

TABLE 11. Silvicultural decision tools.

 

Despite different natural conditions, the volume production was found to be similar in both countries. For two stands (CZ I and HR I) the volume was identical - about 700 m3·ha-1. It is necessary to point out that the analysed stands in CZ were about 20 to 30 years younger than those in HR. According to the existing growth tables for both countries, this involves the most productive stands on top quality soil [19, 20]. The comparable volume yield is also given by the greater number of trees, especially oaks, in the upper layer in CZ, which is related to the silvicultural strategy of the clear-cutting management model applied. Surprising were also the similar oak parameter relationships, which indicate similar dynamics in both countries.

In contrast, the analysis of stand structure confirmed the expected differences between these two countries. While in CZ the poor structure and the tree characteristics observed are given by the clear-cutting management model, the more diverse structure of the forest in HR corresponds to the Croatian model with a multi-layered floodplain forests [12]. Such a model is also closer to the natural conditions in virgin floodplain forests, where the relatively dynamic structure is characterised by a multi-layer distribution of tree species and a distinct diameter differentiation, which particularly applies to the optimum stage [7, 8, 21].

Comparing the individual growth characteristics for crop oaks found in CZ to have up to twice more individuals (130-160 trees per hectare) with smaller spacing than in HR. These oaks, however, had a lower mean volume, shorter crowns and nearly twice smaller crown projection compared with HR oak trees. For instance, Spiecker [6] recommended supporting only about 60 target trees per hectare through crown thinning to optimise the radial increment in oak, which corresponds to the Croatian model. These usually thickest oak trees also have larger crowns (the relationship was confirmed similarly for both countries - see Figure 5 and 6), thus having better prerequisites for fructification [2, 3].Croatian system also provided higher economic benefits. Despite lower cubic volume of oak per ha in HR, we can expect higher economic profit from oak trees in HR due to more valuable assortments (loss in CZ - ca. 22,000 €·ha-1).

The acorn crop values obtained for the analysed stands in both countries (2-17 acorns per square metre) were below the threshold of the mast year (20-50 acorns·m-2) [2, 22]. Years of rich crop are likely to occur more frequently in HR than in CZ [1, 9, 23]. While in mast years the presence of richly and regularly fruiting individuals is not crucial for seeding to be sufficient, in the years with medium and lower crop rates the opposite is true [22, 24]. To this end, the size and quality of the crown and sufficient space for growth, where applicable, are the prerequisites for individual trees to be fruiting richly and on a regular basis [3, 24].

In Croatian management with a multi-layered floodplain forests shelterwood felling in three cuts (preparatory, seeding and final) is used with the regeneration period of 6-10 years, when the average density of oak seedlings and other species is about 40,000-50,000 individuals per ha [12]. However, in CZ shelterwood felling is limited by pure even-aged structure of unprepared adult oak stands which (after opening) causes strong weed expansion and stem sprouting. Dobrovolný [11], however, inventoried (in the southern part of CZ in Židlochovice region with 3355 ha of floodplain forests) in total 8 ha of young pedunculate oak stands (with age 5-7 years) established from natural regeneration with density ranging between 15,000-100,000 individuals per ha. The original mother stands, characterized by lower stock density (that ranged between 0.5-0.8), were harvested by clear-cutting immediately after the acorn fall.

The growth relationships of CZ-HR oaks showed a similar trend (or shape of constructed curves). Thus, the stem and crown characteristics (and also the amount and quality of produced acorns) of new CZ oaks could be probably changed if changing silvicultural system. However, based on the results presented in this paper we can not determine exactly which concrete factor caused the observed differences between HR and CZ. Probably there exists a complex of various factors such as climate, water regime, tree vitality and physiological stress, genetic predispositions, etc. This study should be a start and a challenge for future cooperation and long-term research in this field.

 

CONCLUSIONS

Our results showed elementary differences not only in the forest structure, but also in the management approach in floodplain forests of CZ and HR. While different concepts and stand structures may involve a comparable production level, the approach of clear-cutting management in the Czech Republic brings a range of problems. Despite complex of biotic and abiotic factors and more favourable climatic conditions in HR, the silvicultural system of CZ floodplain forests should be gradually converted to the Croatian model with a multi-layered forest structure, more focused on individual tree growth with high economical value and high reproductive potential. To achieve these goals primarily young and middle-aged stands in CZ should be managed through releasing of crowns of high-quality (and vital) crop trees (60-80 trees per hectare), as well as structuring the stands by preserving the lower tree layers consisting of accompanying tree species.

 

Acknowledgements

The research study was supported by one Project (No. KUS QJ1230330) of the Czech Agency for Agricultural Research and two Projects (No. IGA 84/2013 and No. IGA 33/2014) of Mendel University in Brno.



REFERENCES

  1. KLIMO E, HAGER H, MATIĆ S, ANIĆ I, KULHAVÝ J 2008 Floodplain Forests of Temperate Zone of Europe. Lesnickápráce, Kostelecnad Černýmilesy, 623 p
  2. RÖHRIG E, BARTSCH N 2006 Waldbau auf ökologischer Grundlage [Silviculture on ecological basis]. Stuttgart, Ulmer, Germany, 479 p
  3. MARTINÍK A, DOBROVOLNÝ L, PALÁTOVÁ E 2014 Tree growing space and acorn production of Quercus robur. Dendrobiology 71: 101-108. DOI: https://doi.org/10.12657/denbio.071.010
  4. MATIĆ S, ORŠANIĆ M, BARIČEVIĆ D 1999 Natural regeneration of pedunculate oak in floodplain forests of Croatia. Ekologia 18 (1): 111-119
  5. SORK VL, FLEMING TH, ESTRADA A 1993 Evolutionary ecology of mast-seeding in temperate and tropical oaks (Quercus). Vegetatio 107 (1): 133-147. DOI: https://doi.org/10.1007/BF00052217
  6. SPIECKER H 1991 Zur Steuerung des Dickenwachstums und der Astreinigung von Trauben- und Stieleichen [Influencing of diameter growth and pruning of oaks]. PhD thesis, University in Freiburg, Freiburg, Germany, 150 p
  7. KORPEĹ Š 1991 Pestovanie lesa [Silviculture]. Bratislava, Slovakia, 465 p
  8. REININGER H 2000 Das Plenterprinzip [Principles of selection forests]. Stuttgart, Germany, 238 p
  9. VAŇKOVÁ K 2004 Natural regeneration of oak in floodplain forest. PhDthesis, Mendel University, Brno, Czech Republic, 164 p
  10. HOUŠKOVÁ K, PALÁTOVÁ E, MAUER O 2007. Possibilities and procedures for the natural regeneration of pedunculate oak in south Moravia. In: Proceeding of IUFRO International Conference on Management of Floodplain Forests in Southern Moravia, Židlochovice, Czech Republic, 13-16 May 2000. Židlochovice, Czech Republicpp 89-98
  11. DOBROVOLNÝ L2014 Potential of natural regeneration of Quercus robur in floodplain forests in the southern part of the Czech Republic. Journal of Forest Science 60 (12): 534-539
  12. MATIĆ S 2000 Managing forests of pedunculate oak in different structural and site conditions in Croatia. In: Proceeding of IUFRO International Conference on Management of Floodplain Forests in Southern Moravia, Židlochovice, Czech Republic, 13-16 May 2000. Židlochovice, Czech Republic, pp 55-65
  13. ORŠANIĆ M, DRVODELIĆ D 2007 Natural regeneration of pedunculate oak. Proceeding Conference of Forest Management systems and regeneration of floodplain forest sites, Mendel University of Agriculture and Forestry, Brno, Czech Republic, pp 99-106
  14. PERNAR N, KLIMO E, MATIĆ S, BAKŠIĆ D, LORENCOVÁ H 2009 Different technologies of floodplain forest regeneration from the aspect of soil changes. Journal of Forest Science 55: 357-367
  15. ASSMANN E 1968 Náuka o výnoselesa [Science of forest yield]. Bratislava, Slovakia, 486 p
  16. PETRÁŠ R, PAJTÍK J 1991 Sústavačesko-slovenských objemových tabuliek drevín [Volume table of Czechoslovakia]. Lesnícky časopis 37: 49-56
  17. DÖBBELER H, ALBERT M, SCHMIDT M, NAGEL J, SCHRÖDER J 2011 Programm – Handbuch zur gemeinsamen Version von BWINPro und BWINPro-S [Handbook for BWINPro software]. Göttingen, Dresden, Germany, 126 p
  18. ISTA 2014 ISTA - International rules for seed testing. Zürich, 276 p
  19. ČAVLOVIĆ J, BOŽIĆ M, KOVAČ G 2000 The relationship between the stands elements of pedunculate oak and yield tables in Croatia. In: Proceeding of IUFRO International Conference on Management of Floodplain Forests in Southern Moravia, Židlochovice, Czech Republic, 13-16 May 2000. Židlochovice, Czech Republic, pp 155-168
  20. ČERNÝ M, PAŘEZ J, MALÍK Z1996 Růstové a taxační tabulky hlavních dřevin České republiky [Growth / volume tables of the Czech Rep.]. IFER, Jílové u Prahy, Prague, Czech Republic, 245 p
  21. VRŠKA T, ADAM D, HORT L, ODEHNALOVÁ P, HORAL D, KRÁL K 2006 Developmental dynamics of virgin forest reserves in the Czech Republic- Floodplain forests (Cahnov-Soutok, Ranšpurk, Jiřina). Academia, Prague, Czech Republic, 214 pp.
  22. DEY DC 1995 Acorn production in red oak. Ontario Forest Research Institute, Sault Ste. Marie, Ontario, Canada, 22 p
  23. GRADEČKI-POŠTENJAK M, NOVAK AGBABA S, LICHT R, POSARIĆ D 2011 Dynamics of acorn production and quality of english oak acorn (Quercus robur) in disrupted ecological conditions. Šumar list 135 (13): 169 -181
  24. HEALY WM, LEWIS AM, BOOSE EF 1999 Variation of red oak acorn production. Forest Ecol Manag 116 (1-3):1-11. DOI: https://doi.org/10.1016/S0378-1127(98)00460-5
  25. CLARK PJ, EVANS FC1954 Distance to nearest neighbor as a measure of spatial relationships in populations. Ecology 35 (4): 445-453. DOI: https://doi.org/10.2307/1931034
  26. DEJMAL J 1986 Tabulky pro sortimentaci těžebního fondu [Assortment tables of the Czech Rep.]. University of Agriculture in Brno, Brno, Czech Republic, 25 p
  27. DIACI J, GYOEREK N, GLIHA J, NAGEL T 2008 Response of Quercus robur seedlings to north-south asymmetry of light within gaps in floodplain forests in Slovenia. Ann Forest Sci 65 (1): 105p1–105p8. DOI: https://doi.org/10.1051/forest:20070

© 2017 by the Croatian Forest Research Institute. This is an Open Access paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0).

- Download PDF-

SEEFOR 8 (2): 107-115
Article ID: 163
DOI: https://doi.org/10.15177/seefor.17-15

Original scientific paper

 

Surface Accessibility with Spatial Analysis During Fire Extinguishing Procedures: Example on the Island of Vis


Kruno Lepoglavec1*, Josip Žaček2, Hrvoje Nevečerel1, Ante Seletković3, Zdravko Pandur1, Marin Bačić1


(1) University of Zagreb, Faculty of Forestry, Department of Forest Engineering, Svetošimunska 25, HR-10000 Zagreb, Croatia;
(2) Nova Jošava 34, HR-33515 Orahovica, Croatia;
(3) University of Zagreb, Faculty of Forestry, Department of Forest Inventory and Management, Svetošimunska 25, HR-10000 Zagreb, Croatia

* Correspondence: e-mail:

Citation: LEPOGLAVEC K, ŽAČEK J, NEVEČEREL H, SELETKOVIĆ A, PANDUR Z, BAČIĆ M 2017 Surface Accessibility with Spatial Analysis During Fire Extinguishing Procedures: Example on the Island of Vis. South-east Eur for 8 (2): 107-115. DOI: https://doi.org/10.15177/seefor.17-15

Received: 06 Sep 2016; Revised: 20 Nov 2017; Accepted: 23 Nov 2017; Published online: 4 Dec 2017


Cited by:     Crossref     Google Scholar


Abstract

Background and Purpose: The existing public and forest transport infrastructure (truck forest roads) are permanent objects used when passing through forests. They also serve as a firefighter belt and provide direct access to firefighting vehicles, or are used as the starting point where firefighting teams extinguish fires or move toward remote fires. The paper identifies the existing fire road network (including public roads, forest roads, non-classified roads and fire roads) for access of firefighting vehicles during fire extinguishing interventions.
Material and Methods: An analysis of the intervention rate was conducted on a dispersive sample (35 positions) from two volunteer fire associations (VFA) on the island of Vis. Also, an analysis of the surface availability to fire vehicles concerning the time of departure from the fire station was conducted, as well as the comparison with the Standard time of intervention defined by the regulations on fire department organization in the Republic of Croatia.
Results: For each simulated fire location for intervention of two existing volunteer fire associations: VFA Komiža and VFA Vis, results show that for a few fire locations, despite a smaller distance from the VFA Komiža, a quicker intervention is possible from the VFA Vis (locations 4, 5 and 14), and vice versa (locations 21, 22 and 25). With the use of a New Service Area, tool intervention times regarding different areas were calculated. Intervention times were divided into intervals: <5 min, 5–10 min, 10–15 min, 15–25 min and >25 min. The last two categories of area are beyond reach for firefighters within the Standard time of intervention (15 min) and together they comprise to 27.88% of the total research area.
Conclusions: The results of Closest Facility tool indicate that for the simulated fire position the best/fastest route is not always the shortest one, because of a significant effect of the structural elements of each road, the state of the road and the longitudinal slope of the road itself. One of the possible approaches to gain access to the area for fire-fighting, as well as to prioritize fire roads regarding maintenance/reconstruction is to improve road conditions, and thus increase the average driving speed.

Keywords: fire road infrastructure, fires, access time, GIS analysis 



INTRODUCTION

A forest fire is uncontrollable, destructive movement of fire on the forest surface. It is categorized as a natural disaster and distinguished by type, origin and resulting damage. There are specific needs regarding temperature, pressure and oxygen required for fire, and if one of them is removed, the fire will stop [1]. Forest fires represent a great danger to the forests and forest land in the Republic of Croatia, and are common to the climate in which we live, especially in Dalmatia, Istria, on islands and in Dalmatian Hinterland.

Recently, the danger of forest fires has become extremely high, mostly due to the extensive climate change (long hot summers, warm autumns, strong winds, and long periods with very high temperatures). In Istria and the Croatian Littoral, about 70% of fires emerge in February, March and April, while in Dalmatia most fires occur in July and August. In Croatia, the monitoring and processing of data related to forest fires began in 1955. According to these data, a total of 10,369 fires, or an average of 370 forest fires per year [2] occurred in the period from 1955 to 1984. In the period from 1995 to 2014 in the Republic of Croatia, there were a total of 5,377 fires in forests and other land, and a total of 259,003.17 ha were under fire. In the mentioned twenty-year period, the annual average was 269 fires, with an average annual fire area of 12,950.16 ha [3]. The year 2007 was a record year when 706 forest fires were registered, causing damage to 68,171.00 hectares.

The number of fires in the first half of the year 2017 (or until 15 July) was three times higher than in the year 2016. In the seven coastal counties, there were a total of 642 fires, while in the whole 2016 there were 214 [4]. This area covers 67,397.00 ha.

When we talk about the causes of fires, only about 10% of fires have a known cause, a thunder stroke, while 90% of all fires are a result of accidental or deliberate human action (neglect, burning of agricultural waste, intentional fire, traffic, electricity lines, mines and other) [3]. The emergence of fires in the Dalmatian area is high due to the vegetation cover which consists of coniferous and broadleaved forests, pastures and agricultural land, and due to the neglect of people during agricultural work, soil cleansing and weeding of weeds [5]. A research by Netolicki et al. [6] has shown that the anthropogenic influence is considered to be the major factor in the outbreak of fires. High influence also lies in terrain morphology, geological substrate and soil. As Rosavec [7] points out, the higher probability of fires and the amount of burned surface can be determined by the condition of the vegetation and the climate. Martinović [8] points out that in the USA the most considerable damage is caused by forest fires, similar to those in our karst forest ecosystems, and points to the fact that attention should be paid to the pedological conditions of forest fires.

There are two groups of fire protection measures, both preventive and curative. Preventive measures are used to prevent or reduce the possibility of fires, while curative include the process of extinguishing and repairing the burnt area. Exceptional measures of both preventive and curative protection are fire roads. During fires, fire roads serve as a firefighting belt, provide access to firefighting vehicles, emergency vehicles and vehicles for the transportation of personnel and equipment, and can also serve as a place for firefighters to wait for the future fires, as well as places for pre-fire and anti-fire ignitions [9].

In this paper, the analyses are based on the use of firefighting vehicles for firefighting interventions, and the term “intervention” is considered as a movement of a fully qualified vehicle and equipment until the vehicle reaches the endpoint of the fire road. Of course, the intervention can be considered to consist of unified operations from the call itself up to access and shutdown or localisation of the fire.

Since time is the most important factor for reaching a fire, every efficient firefighting system, due to its rapid localisation, requires well-planned intervention, an appropriate risk assessment and management system, comprehensive training, and quick implementation of the above-mentioned steps through an application. Technology with a growing frequency of use in optimising this system is the Geographic Information System (GIS) [10-12]. Every day firefighters are faced with ever-increasing demands for work, so they have been forced to implement state-of-the-art tools, techniques and methods [13]. The imperative of all firefighting interventions is the speed and accuracy of the reaction. In this context, using GIS technologies enables us, thanks to implemented algorithms, to eliminate possible human errors when selecting a route, thus significantly shortening the time of intervention [13]. The most common data layers used by fire departments are streets, parcels, hydrants, public networks, rivers and lakes, business buildings, police and fire stations, schools and hospitals, satellite imagery and previous fire locations [11, 14].

 

MATERIAL AND METHODS

The main tool for conducting the analysis is the Network Analyst. It is a powerful ArcGIS extension and enables analysis based on topologically accurate traffic data [13]. It is useful in firefighting because it enables: to define the fire department closest to the fire area, estimate the travel time, select a new potential location, find the fastest route, nearest fire station, or define the optimal deployment of the existing fire departments. Three components are important when selecting a faster firefighting intervention: the location of the fire, the location of the fire department and the distance from the unit to the fire location.

The establishing of the database relies on the existing digitalized network of roads and on adding newly established traffic infrastructure (Figure 1), in which all “controversial” road segments are corrected directly on the field and recorded using GPSMAP 62S GPS, brand Garmin. In this paper cartographic background was used which was made in the transverse Mercator projection and by the reference coordinate system HTRS96 (Croatian Terrestrial Coordinate System at epoch 1995.55).

 

FIGURE 1. The method of digitizing a network of trafficable roads on digital orthophoto (DOP).

 

The control time of vehicle arrival at the test sites, i.e. the average travel time of the fire truck on particular road segments, and the trace recording were done by a GPS device mounted on the Mercedes Atego 1528 fire truck. This resulted in average speeds of the vehicle in all the routes used in the analyses, shown in different colours for a certain average driving speed (Figure 2). By calculating the length of each road segment and the required transition time, all the parameters necessary for calculating the fastest possible path to simulated fire positions can be obtained.

 

FIGURE 2. Transport infrastructure network with average driving speed.

 

Based on such structured data layer, an analysis in Network Analyst can begin. With the aid of this tool, the fastest/closest firefighter unit to the fire position (on a dispersive sample) is defined. The tool also finds the fastest route and estimates the travel time to the site of intervention. Within Network Analyst tools, Closest Facility tool is used, which is based on Dijkstra’s algorithm1 or shortest path algorithm. The algorithm breaks the network into nodes, and the routes that link them are visualised by the vector line data with the attribute values. Additionally, each line between two nodes has a related value (cost or distance) that needs to be overcome in order to reach the destination node or point [15]. An important factor when choosing a route is not only speed but also traffic conditions on the road network, which in this case are the average driving speeds on certain road segments.

The model created for this research within Network Analyst tools, a tool entitled New Service Area was used. This tool gives us an output polygon that shows the area of a given firefighting station’s intervention period through time aspect and distance. Also, tools such as Select, Clip, Merge, and Erase have also been used to calculate the availability of different surface areas at a certain time.

The aim of the analysis is to identify the location that covers the largest area and that makes responding to fires in standard intervention time possible. Standard intervention time is defined as the standard time set by the regulations on fire department organization in the Republic of Croatia, in which Article 19 states: “The distribution of fire brigade units on the territory of the Republic of Croatia should be such that the arrival of the fire brigade to intervene to the furthest place of the protected area is set to a fifteen-minute limit.”.

Research Area

Forest administration (FA) office Split is one of the 16 administrations included within the Croatian Forests Ltd company. This FA manages forests between the Pag Bridge and Prevlaka, on the territory of four counties: Zadar, Šibenik-Knin, Split-Dalmatia and Dubrovnik-Neretva. The total area covered by the FA is 563,804.38 ha, which is also the largest area covered by one administration office. Out of the total forest areas covered by this management, 444,175.16 ha are covered forest areas, 105,825.20 ha are uncovered forest areas and 13,804.02 ha are barren forest areas [17].

 

FIGURE 3. Research area – the island of Vis.

 

FA Split includes 986 islands, five nature parks and four national parks. In the coastal area, FA Split manages species preservation, forest protection, planting and other activities. With the rise of summer heat, the fear of possible upcoming fires grows, since fires are the greatest enemies of forest land. In the karst area, the greatest threat to forests are forests fires, so a lot of money is invested in the preventive protection of forests. Anti-fire prevention measures include the organisation of observatory firefighting service, the construction and maintenance of observation posts, the construction and maintenance of forest fire roads, the placing of warning signs and the preservation of forests. A big problem is that volunteer fire associations on the Adriatic islands and coastal areas have lately been in great trouble, quite often at the border of existence. The reason for this is primarily that a small number of young people are included in the associations, and that there is a growing lack of interest, insufficient equipment, obsolete equipment, and inability to acquire new equipment [5].

The island of Vis has been selected for research for a number of reasons, primarily due to the existence of two active voluntary firefighter associations, the existence of a large number of different road categories and their conditions (level of damage) with a total of 208.50 km in length, great distance of the island from the mainland (about 45 km), and the fact that it is wholly unavailable for quick firefighting interventions from the air, so all fire protection and intervention depend on the existing roads.

The surface of the island is 90.3 km2, the total length of the indented and quite inaccessible coast is 77 km. The island of Vis, according to the WGS84 geographic projection, is located between 16°02’22’’E 43°00’13’’N and 16°16’13’E 43°04’53’’N.

 

RESULTS

The tools used in this research enabled us to create a supplemented road cadastre that was the input for all the necessary analysis foreseen in this research. The total length of roads that can be used in firefighting interventions is 208.50 kilometres, and the existing roads are divided into 552 segments of the researched road network with assigned average vehicle driving speeds. The road network designated for firefighting interventions is divided into segments defined by nodes (intersection points), i.e. intersections and road endings that according to the tool use and represent a mandatory intercept.

By simulation/random selection, 35 points have been set in the entire research area that represents areas of eventual fire (Figure 4). The points were determined by order of 1 to 35, and a dispersed pattern was set. This analysis would point out that the closest route is not always the shortest one during the intervention due to different conditions of the roadway. The tool has proved to be efficient for making sensible objective decisions in the logistics of the fire extinguishing system.

 

FIGURE 4. Spatial positions of simulated fires.

 

To confirm a dispersive pattern of the simulated fires, a statistical analysis was carried out using the Average Network Neighbor tool. Given the value of Nearest Neighbor Ratio of 1.532692, a p-value of 0.000000 and a z-score of 6.028936, there is a less than 1% likelihood that this dispersed pattern could be the result of random chance, proving the dispersed layout of the locations (Figure 5).

 

FIGURE 5. The results of testing the selected locations using the Average Network Neighbor tool.

 

For each simulated fire location, an analysis was conducted regarding the required time and length of the access route for intervention along with the two locations of existing volunteer fire associations: VFA Komiža and VFA Vis (Table 1). Time was displayed in minutes and distance in kilometres. The results show that for a few fire locations, despite a smaller distance from the VFA Komiža, a quicker intervention is possible from the VFA Vis (locations 4, 5 and 14), and vice versa (locations 21, 22 and 25). The reason for this is terrain configuration and the degree of road damages, which cause a decrease in the average driving speed of firefighter trucks on certain road segments, and therefore it takes longer for the firefighters to arrive at the intervention site.

 

TABLE 2. Intervention time and route distance for the simulated fire positions.

 

With the use of a New Service Area, tool intervention times regarding different areas were calculated. Intervention times were divided into intervals: <5 min, 5–10 min, 10–15 min, 15–25 min and >25 min, so that it is possible to differentiate areas accessible within the standard time of 15 minutes. The last two categories (15–25 min and >25 min) shown in Figure 6 are beyond reach for firefighters within the standard time and together they comprise to 2,530.62 ha which is 27.88% of the total research area.

 

FIGURE 6. Firefighters’ access within different time intervals.

 

Areas accessible at different time intervals were calculated for each of the volunteer fire associations separately to determine the area coverage of a particular firefighter unit. The results show that areas accessible to firefighters within 5 minutes significantly differ between these two volunteer fire associations. The area accessible within 5 minutes to VFA Vis is twice the size of the area accessible to VFA Komiža in the given period.

This difference increases in favor of VFA Vis by increasing the time of intervention and within the standard time where almost ¾ of the areas’ coverage/accessibility is in favor of VFA Vis, as opposed to the ¼ surface accessibility of VFA Komiža (Figure 7). It is also noticeable that after a time interval of 15 minutes, area coverage of VFA Vis enlarges, while of VFA Komiža it slightly decreases. Namely, during a time of intervention of more than 15 minutes, more rapid interventions are those of VFA Vis, regardless of the greater geometric distance in the case of VFA Komiža, all due to better linkage and driving speed from the direction of VFA Vis.

 

FIGURE 7. Firefighters’ access within different time intervals for a particular volunteer fire association.

 

DISCUSSION AND CONCLUSIONS

The only efficient way to minimize damages caused by forest fires is the early detection of forest fires and fast and appropriate reaction, apart from applying preventive measures. Considerable efforts are therefore made to achieve early forest fire detection, which is traditionally based on human surveillance [18]. Therefore, it is crucial to determine the optimum route that minimizes the travel time of the initial response team from fire headquarters to fire areas using firefighting trucks [19]. It is essential to pay close attention to a well-developed road network that allows access to fires on islands that are far off the mainland and where rapid air intervention is not possible, especially in the summer months when these areas have a high risk of fire due to long dry periods without rainfall and highly flammable plant species.

To maximize the existing traffic infrastructure for firefighting, it is important to determine the state of the road, its trafficability and the possible driving speed of firefighting vehicles. With the knowledge of all these details and GIS, it is possible to find the best and the fastest solution for individual firefighting interventions. Firefighting units have several GIS tools available for the analysis of intervention speed, which, as shown in this paper, can be categorised within the Network Analyst tools. It is also possible to conduct other analysis such as complex modelling of various hazard indexes, the degree of fire risk, fire susceptibility, topography and weather conditions, simpler visibility analysis, selection of optimal location of fire lookout towers and fire stations, determination of intervention location, etc. [13].

The results obtained with the Closest Facility tool indicate that for the simulated fire position the best/fastest route is not always the shortest one, because of a significant effect of the structural elements of each road, the state of the road and the longitudinal slope of the road itself. In addition to the variables used in this paper (average driving speed and road length), it is possible to define obstacles and constraints that block or hinder traffic on certain road segments [20]. For analysis of obstacles and changes in road conditions, it is necessary to carry out real-time analysis whereby the previously formed databases would obtain current road conditions, which would result in a change of intervention route [21, 22].

The New Service Area tool showed that approximately ¼ of the island’s surface is unavailable to firefighters within a 15-minute standard time. The cartographic presentation gives us guidelines for reconstruction or maintenance of the existing roads, and thus a total increase in the average driving speed of firefighting vehicles would be possible. The areas shown in Figure 5 are a good indicator of areas where new roads need to be built to shorten the drive, all for comprehensive protection and possible interventions in the entire research area.

An important indicator of the conducted analysis is that much larger surface area is available to the VFA Vis within the standard time (Figure 7), because there are higher road categories in the vicinity of the city of Vis where the VFA is located, which then allows faster movement of firefighters due to improved traffic conditions and two traffic lanes.

The density of roads suitable for firefighting interventions is 23 m·ha-1. The established road density was shown to be insufficient, which prevents timely intervention on all the areas covered by this research. This is also contributed by the poor state of the upper rod layer on a large number of roads, which decreases the driving speed on both asphalt and macadam roads. One of the important problems identified is the position of volunteer fire associations located in the coastal part of the island in the centre of the cities of Vis and Komiža. They are not placed in ideal positions regarding road layout, terrain configuration and island indentation. This case shows the need to set up new firefighting stations or to seasonally displace them for better efficiency and better protection [23].

Taking into account all the given results in this paper, the time of intervention can be reduced by appropriate planning and realization of the proposed measures. The existing network of roads needs to be improved in qualitative and quantitative terms, and the available VFAs should be brought closer to possible interventions. This research has also shown the need to redefine the existing knowledge of the optimal density of roads in the island karst area because in the case of a fire the time of approach determines the success of the entire intervention system. It is not possible to propose a final solution by this research, but many questions arise that open the way for new research, oriented towards conscious and rational surface management where there is an obvious high risk of fire.


REFERENCES

  1. PIČMAN D 2011 Interna skripta iz nastavnog predmeta Šumske protupožarne prometnice. University of Zagreb, Faculty of Forestry, Zagreb, Croatia, pp 1-63 (in Croatian)
  2. ŽAČEK J 2017 Fire fighting infrastructure analysis of island Vis. Diploma thesis, University of Zagreb, Faculty of Forestry, Zagreb, Croatia, ­42 p (in Croatian)
  3. BAKŠIĆ N, VUČETIĆ M, ŠPANJOL Ž 2015 A potential risk of fire on open space in the Republic of Croatia. Vatrogastvo i upravljanje požarima (Fire fighting and management) 2: 30-40 (in Croatian with English summary)
  4. GOVERNMENT OF THE REPUBLIC OF CROATIA 2017 URL: https://vlada.gov.hr/vijesti/sjednica-vlade-broj-pozara-u-2017-je-u-porastu-identificirat-cemo-preventivne-mjere-kako-bismo-smanjili-rizike/22110 (2 August 2017)
  5. NODILO J 2003 Fires on open spaces, islands and the littoral - an accident or a logical sequence of events? Šumarski list 127 (3-4): 171-176 (in Croatian with English summary)
  6. NETOLICKI A, BLAŽEVIĆ T, ANTOLOVIĆ A 2012 Multicriteria Analysis of Fire Risk in the Split-Dalmatia County. Cartography and geoinformation 11 (17): 4-24
  7. ROSAVEC R 2010 Relation between climate factors and the flammability of some Mediterranean species in forest fires. PhD thesis, University of Zagreb, Faculty of Forestry, Zagreb, Croatia, 175 p
  8. MARTINOVIĆ J 2003 Management of forest soils in Croatia. Forest research institute Jastrebarsko and Croatian Forest Ltd., Jastrebarsko, Zagreb, Croatia, 521 p
  9. FRANULOVIĆ K 2014 Protupožarne prometnice na kršu. Savjetodavna služba, URL: dostupno na: http://www.savjetodavna.hr/savjeti/558/563/protupozarne-prometnice-na-krsu/ (15 August 2017)
  10. MALOY MA, DEAN DJ 2001 An accuracy assessment of various GIS-based viewshed delineation techniques. Photogramm Eng Rem S 67 (11): 1293-1298
  11. VASILAKOS C, KALABOKIDIS K, HATZOPOULOS J, KALLOS G, MATSINOS Y 2007 Integrating new methods and tools in fire danger rating. Int J Wildland Fire 16 (3): 306-316. DOI: https://doi.org/10.1071/WF05091
  12. PATAH NA, MANSOR S, MISPAN MR 2001 An application of remote sensing and geographic information system for forest fire risk mapping. Malaysian Centre for Remote Sensing (MACRES), pp 54-67
  13. ROLAND V, MARIĆ I, MILOŠEVIĆ R 2015 Application of GIS technology in firefighting. Vatrogastvo i upravljanje požarima (Fire fighting and management), 1 (5): 57-71 (in Croatian with English summary)
  14. CHUVIECO E, SALAS J 1996 Mapping the spatial distribution of forest fire danger using GIS. Int J Geograph Inf Sci 10 (3): 333-345. DOI: https://doi.org/10.1080/02693799608902082
  15. KARADIMAS N V, KOLOKATHI M, DEFTERAIOU G, LOUMOS V 2007 Ant colony system vs ArcGIS network analyst: The case of municipal solid waste collection. In: 5th WSEAS International Conference on Environment, Ecosystems and Development, Tenerife, Spain, pp 128-134
  16. ODOBAŠIĆ D 2009 Specifikacija i implementacija mrežnih modela prostornih podataka. Diploma thesis, University of Zagreb, Faculty of Geodesy, Zagreb, Croatia, 82 p
  17. DRAGIČEVIĆ JS 2007 Gospodarenje šumama u Hrvatskoj i na primorskom području. Građevinar 59: 447-453
  18. ŠĆEPANOVIĆ J, BUČAN P, KOVAČEVIĆ I 2012 Analysis of intervention fire extinguishing “DES” Split. Vatrogastvo i upravljanje požarima (Fire fighting and management) 2 (2): 67-80 (in Croatian with English summary)
  19. AKAY AE, WING MG, SIVRIKAYA F, SAKAR D 2012 A GIS-based decision support system for determining the shortest and safest route to forest fires: a case study in Mediterranean Region of Turkey. Environ Monit Assess 184 (3): 1391-1407. DOI: https//doi.org/10.1007/s10661-011-2049-z
  20. KERAMITSOGLOU I, KIRANOUDIS CT, SARIMVEIS H, SIFAKIS N 2004 A multidisciplinary decision support system for forest fire crisis management. Environ Manage 33 (2): 212-225. DOI: https://doi.org/1007/s00267-003-0092-4
  21. GENDREAU M, LAPORTE G, SEMET F 2001 Adynamic model and parallel tabu search heuristic for real-time ambulance relocation. Parallel Comput 27 (12): 1641-1653. DOI: https://doi.org/10.1016/S0167-8191(01)00103-X
  22. GHIANI G, GUERRIERO F, LAPORTE G, MUSMANNO R 2003 Real-time vehicle routing: Solution concepts, algorithms and parallel computing strategies. Eur J Oper Res 151 (1): 1-11. DOI: https://doi.org/10.1016/S0377-2217(02)00915-3
  23. DRŽAIĆ D, KAVRAN M, ANTOLOVIĆ A 2014 Determination of Appropriate Locations for Seasonal Dislocation of Fire Stations in the Šibenik-Knin County Based on Road Network Analysis. Cartography and geoinformation 22 (13): 20-31


© 2017 by the Croatian Forest Research Institute. This is an Open Access paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0).

- Download PDF-

SEEFOR 8 (2): 117-125
Article ID: 171
DOI: https://doi.org/10.15177/seefor.17-16

Original scientific paper

 

The Evaluation of Photogrammetry-Based DSM from Low-Cost UAV by LiDAR-Based DSM


Mateo Gašparović1, Ante Seletković2, Alen Berta3, Ivan Balenović4*


(1) University of Zagreb, Faculty of Geodesy, Chair of Photogrammetry and Remote Sensing, Kačićeva 26, HR-10000 Zagreb, Croatia;
(2) University of Zagreb, Faculty of Forestry, Department of Forest Inventory and Management, Svetošimunska 25, HR-10000 Zagreb, Croatia;
(3) Oikon Ltd. Institute of Applied Ecology, Department of Natural Resources Management, Trg Senjskih Uskoka 1-2, HR-10000 Zagreb, Croatia;
(4) Croatian Forest Research Institute, Division for Forest Management and Forestry Economics, Trnjanska cesta 35, HR-10000 Zagreb, Croatia

* Correspondence: e-mail:

Citation: GAŠPAROVIĆ M, SELETKOVIĆ A, BERTA A, BALENOVIĆ I 2017 The Evaluation of Photogrammetry-Based DSM from Low-Cost UAV by LiDAR-Based DSM. South-east Eur for 8 (2): 117-125. DOI: https://doi.org/10.15177/seefor.17-16

Received: 14 Oct 2017; Revised: 25 Nov 2017; Accepted: 29 Nov 2017; Published online: 11 Dec 2017


Cited by:     Crossref     Google Scholar 


Abstract

Background and Purpose: Unmanned aerial vehicles (UAVs) are flexible to solve various surveying tasks which make them useful in many disciplines, including forestry. The main goal of this research is to evaluate the quality of photogrammetry-based digital surface model (DSM) from low-cost UAV’s images collected in non-optimal weather (windy and cloudy weather) and environmental (inaccessibility for regular spatial distribution of ground control points - GCPs) conditions.
Materials and Methods: The UAV-based DSMs without (DSMP) and with using GCPs (DSMP-GCP) were generated. The vertical agreement assessment of the UAV-based DSMs was conducted by comparing elevations of 60 checkpoints of a regular 100 m sampling grid obtained from LiDAR-based DSM (DSML) with the elevations of planimetrically corresponding points obtained from DSMP and DSMP-GCP. Due to the non-normal distribution of residuals (vertical differences between UAV- and LiDAR-based DSMs), a vertical agreement was assessed by using robust measures: median, normalised median absolute deviation (NMAD), 68.3% quantile and 95% quantile.
Results: As expected, DSMP-GCP shows higher accuracy, i.e. higher vertical agreement with DSML than DSMP. The median, NMAD, 68.3% quantile, 95% quantile and RMSE* (without outliers) values for DSMP are 2.23 m, 3.22 m, 4.34 m, 15.04 m and 5.10 m, respectively, whereas for DSMP-GCP amount to -1.33 m, 2.77 m,  0.11 m, 8.15 m and 3.54 m, respectively.
Conclusions: The obtained results confirmed great potential of images obtained by low-cost UAV for forestry applications, even if they are surveyed in non-optimal weather and environmental conditions. This could be of importance for cases when urgent UAV surveys are needed (e.g. detection and estimation of forest damage) which do not allow careful and longer survey planning. The vertical agreement assessment of UAV-based DSMs with LiDAR-based DSM confirmed the importance of GCPs for image orientation and DSM generation. Namely, a considerable improvement in vertical accuracy of UAV-based DSMs was observed when GCPs were used.

Keywords: stereo photogrammetry, unmanned aerial vehicle (UAV), digital surface model (DSM), Structure from Motion (SfM), light detection and ranging (LiDAR), vertical agreement assessment, forest inventory



INTRODUCTION

Today we are witnessing the growing use of unmanned aerial vehicles (UAVs) for monitoring purposes. Potential applications of UAVs can be found in agricultural, forestry, and environmental sciences; surveillance, and reconnaissance; aerial monitoring in engineering; cultural heritage; and traditional surveying, conventional mapping and photogrammetry, and cadastral applications [1]. Due to various construction solutions UAVs are flexibile to solve various surveying tasks. Compared to the classical terrestrial survey, UAVs are capable to cover considerably larger areas in short time period, as well as to survey distant or inaccessible areas (e.g. distant forest and mined areas) and objects (e.g. high buildings). The flexibility of photogrammetric surveying methods along with the selection of the adequate cameras and lenses results in adaption of the measuring platform (UAV) to the needs of the tasks. Furthermore, UAVs have a capability of an autonomous recording, and hence they are becoming independent devices for gathering a large number of high-quality data of the field or object with appropriate accuracy.

Recently, comprehensive reviews on applications of UAVs in forestry have been provided by Tang and Shao [2] and Torresan et al. [3]. In general, the common UAVs applications in forestry are related to monitoring of forest health and disturbances [4-6], forest inventory [7, 8], forest cover mapping [9], etc. Digital surface model (DSM), which is one of the main photogrammetric products of UAV surveys, has great application in forest inventory. By subtracting available digital terrain model (DTM), which presents terrain surface, from DSM, which presents forest surface, a canopy height model (CHM) is generated. DTMs are nowadays commonly generated using airborne laser scanning (ALS) technology based on light detection and ranging (LiDAR) or airborne digital photogrammetry [10]. From CHMs various metrics can be derived which are then used for estimation of various tree [11] and stand variables [7, 12]. The Structure from Motion (SfM) algorithm has been suggested for DSM generation by many authors [13-15]. Camera calibration and image phototriangulation process are initially performed to generate accurate DSM or digital terrain model (DTM) [16]. Camera calibration method and the algorithm for the precise elimination of lens distortion on digital cameras was developed by Gašparović and Gajski [17]. Continuing the research Gašparović and Gajski [18] presented a new method of two-step camera calibration for micro UAVs.

Methods for producing photogrammetric DSMs without using ground control points (GCPs) were presented in several studies [19-21]. To obtain external orientation parameters, Chikhradze [19] used single-frequency Global Navigation Satellite Systems (GNSS) receivers, while Vander et al. [20] and Fazeli et al. [21] used dual-frequency differential GNSS. Furthermore, Gimbal influence on the stability of exterior orientation parameters of UAV images was examined in study by Gašparović and Jurjević [22].

The DSMs generated from airborne digital stereo images were evaluated in many studies [e.g. 23-25] which revealed that many factors may influence on their quality, especially in complex forest structure. The research on DSM quality obtained from UAV images are still lacking (especially in South-east European region), but it can be assumed that apart from technical characteristics related to UAV (e.g. camera quality, GNSS precision) similar factors (e.g. image quality, algorithm for image processing, weather conditions, forest structure, etc.) are present.

The main goal of this research is to evaluate the quality of photogrammetry-based DSM from low-cost UAV’s images collected in non-optimal weather (windy and cloudy weather) and environmental (inaccessibility for regular spatial distribution of GCPs) conditions. Namely, urgent cases (e.g. detection and estimation of forest damage) sometimes require rapid and immediate reaction when data acquisitions have to be done in non-optimal weather conditions during the survey. Furthermore, in dense forests it is very difficult to find a place for GCPs, especially to obtain the regular spatial distribution of GCPs which will provide the most accurate orientation of images. Therefore, vertical agreement assessment of UAV-based DSMs generated without and with using GCPs was evaluated with LiDAR-based DSM in this study.

 

MATERIALS AND METHODS

Study Area

The research was conducted in the lowland forest complex of Pokupsko Basin located 35 km southwest of Zagreb, Central Croatia (Figure 1). The study area (77.39 ha) encompasses two 45-year-old mixed forest stands (subcompartments 36a and 37a, management unit “Jastrebarski lugovi”) dominated by pedunculate oak (Quercus robur L.) accompanied by black alder (Alnus glutinosa (L.) Geartn.), common hornbeam (Carpinus betulus L.), and narrow-leaved ash (Fraxinus angustifolia Vahl.), and with the Corylus avellana L. and Crataegus monogyna Jacq. in the understorey. The study area is flat, with ground elevations ranging from 108 to 113 m a.s.l.

 

FIGURE 1. (a) Location of the study area; (b) Study area with 7 GCPs and 60 checkpoints of the regular 100 m sampling grid (background: satellite image WorldView-3, "true colour" composite (5-3-2), sensing date: 12 June 2017).

 

UAV-Based Canopy Digital Surface Models

The UAV images were acquired using the DJI Phantom 4 Pro UAV with FC6310 camera (Table 1) on 14 September 2017. The average flying height was 200 m above ground level. The study area was covered by 488 RGB images with the ground sampling distance (GSD) of approximately 5 cm. The images were collected in 11 flight lines with endlap of 90% and sidelap of 80%. Weather conditions during UAV survey were not suitable (non-optimal) due to windy and cloudy weather.

 

TABLE 1. Characteristics of FC6310 camera.

 

Before the UAV survey, seven ground control points (GCPs) were placed and measured in the study area (Figure 1). The GCPs’ positions (x, y, z coordinates) were measured using the Trimble GNSS receiver connected with the Croatian Positioning System (CROPOS) which enables to obtain both horizontal and vertical positional accuracy from 2 to 5 cm (CROPOS - Users’ Manual). Due to dense forest and mostly invisible ground from the air, it was not possible to provide (set up) the regular spatial distribution of GCPs over the entire study area which enables the most accurate orientation of images [26]. Therefore, GCPs were set up and measured on the forest roads from where they can be easily detected on UAV images (Figure 1). 

From the collected UAV images, two DSMs were generated. First DSM was generated without using GCPs. This means that DSM was generated from UAV images whose orientation was based on a priori exterior orientation parameters (EOPs) only. A priori EOPs were measured during flight in metadata files of each image by GNSS. Firstly, tie-points on all images were automatically determined using the Structure from Motion (SfM) algorithm. Image coordinates of tie-points and a priori EOPs were then used for photo-triangulation with self-calibration. By automatic correlation of oriented images, the point cloud was obtained and then used to generate raster DSM (hereinafter referred to as DSMP) with a spatial resolution of 0.5 m.       

To generate the second DSM, the classic image photo-triangulation method based on tie-points and GCPs was used. Tie-points on all images, as in the previous case, were automatically determined using SfM algorithm. Photo-triangulation with self-calibration was based on image coordinates of tie-points and GCPs, and GCPs’ coordinates in the terrestrial coordinate system. A priori EOPs were not used in this case. A raster DSM (hereinafter referred to as DSMP-GCP) with a spatial resolution of 0.5 m was generated from the point cloud obtained by automatic correlation of oriented images.          

The whole procedure of image orientation and DSMs generation was performed using Agisoft PhotoScan software (version 1.2.6, 64 bit).


LiDAR-Based Canopy Digital Surface Model

A raster LiDAR-based DSM (hereinafter referred to as DSML) with a spatial resolution of 0.5 m was provided by Hrvatske vode Ltd. (Zagreb, Croatia). Table 2 provides an overview of LiDAR sensor and data characteristics used for DSML generation. The resulting point densities (11.59 points·m-2) and the stated horizontal (0.15 m) and vertical (0.08 m) accuracies were based on a considerably larger area (which included and non-forested areas as well) than the one considered in this study. DSML was generated from returns classified as “first return” and “only return”. DSML was used as reference data for vertical agreement assessment of UAV-based DSMs (DSMP and DSMP-GCP). Due to its high accuracy, the LiDAR data were often used as reference data for evaluation of UAV data [27-29].

 

TABLE 2. LiDAR sensor and data characteristics.

 

Vertical Agreement Assessment

The vertical agreement assessment of the UAV-based DSMs was conducted by comparing elevations of 60 checkpoints of a regular 100 m grid obtained from DSML with the elevations of planimetrically corresponding points obtained from DSMP and DSMP-GCP. Prior to defining measures for agreement assessment, the normality of residuals (vertical errors between UAV- and LiDAR-based DSMs) distribution was analyzed using: (a) histograms with a superimposed curve indicating normal distribution, (b) Shapiro-Wilk test [30, 31], and (c) normal Q-Q plots (Figure 2). All performed tests revealed non-normal distribution of vertical errors for both UAV-based DSMs. Consequently, the following robust measures suggested by Höhle and Höhle [10] were used for vertical agreement assessment: median, normalised median absolute deviation (NMAD), 68.3% quantile and 95% quantile. Additionally, root mean square errors before (RMSE) and after removing outliers (RMSE*) were calculated. The equations for all measures, as well as for the threshold for outliers can be found in Höhle and Höhle [10]. The statistical analyses were performed using the STATISTICA software (version 11) [32] and R programming language (version 3.3.3) [33].  

 

FIGURE 2. Normality test of residuals (vertical errors between UAV- and LiDAR-based DSMs): (a) and (b) histograms with a superimposed curve indicating normal distribution with accompanied results of the Shapiro-Wilk test; (c) and (d) indicate normal Q-Q plots.

 

To support statistical analyses, the visual assessment of UAV- and LiDAR-based DSMs, as well as the visual assessment of difference raster models was performed. Difference raster models were generated by subtracting LiDAR-based from UAV-based DSMs. Both, difference raster model generation and its visualization were conducted using Global Mapper (version 19) [34] and QGIS (version 2.18) [35] software.

 

RESULTS AND DISCUSSION

According to the described methods, DSMP (Figure 3a) and DSMP-GCP (Figure 3b) were generated. Detailed information on DSMs processing is presented in Table 3. It can be seen that computer processing time for both DSMs is almost the same, whereas the time spent on manual work is considerably greater for DSMP-GCP generation (30 min) than for DSMP generation (10 min). Namely, during the DSMP-GCP generation, most of the time (≈20 min) was spent on the manual detection of the GCPs on images, while for the DSMP generation the UAV images were orientated without using GCPs.

 

TABLE 3. Information on UAV image orientation and DSMs processing.

 

FIGURE 3. (a) UAV-based digital surface model generated without using GCPs (DSMP); (b) UAV-based digital surface model generated using GCPs (DSMP-GCP); (c) Vertical profile throughout the exemplary area marked with black line on figures (a) and (b) (DSML - LiDAR-based digital surface model; DTML - LiDAR-based digital terrain model).

 

The results of the vertical agreement assessment of the UAV-based DSMs (DSMP and DSMP-GCP) with DSML conducted on 60 checkpoints of the regular 100 m sample grid are shown in Table 4. When comparing UAV-based DSMs with DSML, it is necessary to have in mind that between the acquisition of LiDAR and UAV data is a time gap of one year which corresponds with one vegetation and subsequently with annual height increment. According to the internal database (unpublished material) of Croatian Forest Research Institute, annual height increment for the forest of the study area ranges from 0.2 m to 0.45 m depending on tree species. As expected, DSMP-GCP shows higher accuracy, i.e. higher vertical agreement with DSML than DSMP. Namely, the horizontal accuracy (RMSEXY) of DSMP assessed with 7 GCPs (which were not used in its generation) is 5.67 m (Table 3). Since such horizontal errors may produce greater vertical errors [36], especially for surfaces with great variations in height on a small area (e.g. forest) [25, 37], the lower vertical agreement of DSMP with DSML is understandable. This is especially evident in Figure 3c, which shows a comparison of DSMs’ profiles through the exemplary area. By observing profiles at greater peaks, it can be seen that DSMP-GCP profile follows the DSML profile, whereas for DSMP profile the horizontal displacement of 5-10 m compared to DSML profile can be observed. The improvement in vertical agreement of UAV-based DSMs with DSML when GCPs are used can be observed visually on difference models (Figure 4). Similarly, when comparing two DSMs derived from WorldView-2 images, Hobi and Ginzler [38] found clear improvement of the DSM’s vertical accuracy when GCPs were used.

 

TABLE 4.The vertical agreement assessment of the UAV-based DSMs with LiDAR-based DSM.

 

FIGURE 4. (a) Difference model generated by subtracting DSML from DSMP; (b) Difference model generated by subtracting DSML from DSMP-GCP.

 

Furthermore, Figure 3c shows that DSML provides the highest discrimination of vertical forest structure clearly describing very steep variations in height (e.g. small gaps in the forest canopy, forest road). On the contrary, the profiles of both UAV-based DSMs are considerably smoother. Only bigger gaps in the forest canopy and a forest road (Figure 3a and 3b) can be detected, but in both cases, the vertical profiles of UAV-based DSMs do not reach the ground elevations. This is reasonable because LiDAR is an active sensor whose beams can penetrate through smaller gaps in the forest canopy and reach the ground, whereas the digital camera of UAV system used in this research (Table 1) is a passive optical sensor whose signal can characterize only the canopy surface [39].

Besides the technical limitations of low-cost UAV (e.g. camera quality, GNSS precision) used in this study and non-regular spatial distribution of GCPs, it can be suggested that the weather conditions (windy and cloudy weather) during UAV survey influenced image quality to a certain extent and consequently DSMs quality. The uncertainties are larger due to the complexity of the forest environment (e.g. moving trees, occlusions, shadows, images radiometric quality, etc.), which seriously affect the image matching procedure, and thus DSM quality [23, 25, 40, 41].

 

CONCLUSIONS

This research confirmed great potential of images obtained by low-cost UAV for forestry applications, even if they are surveyed in non-optimal weather (windy and cloudy weather) and environmental (inaccessibility for regular spatial distribution of GCPs) conditions. This could be of importance for cases when urgent UAV surveys are needed (e.g. detection and estimation of forest damage) which do not allow careful and longer survey planning. The vertical agreement assessment of UAV-based DSMs with LiDAR-based DSM confirmed the importance of GCPs for image orientation and DSM generation. Namely, a considerable improvement in vertical accuracy of UAV-based DSMs was observed when GCPs were used. While DSMs generated without GCPs can be used for visualisation and monitoring purposes, DSMs generated with GCPs have potential to be used in forest inventory. To confirm this, further research should focus on estimating the accuracy of tree and stand attributes.

 

Acknowledgments

This research has been supported by the Croatian Science Foundation under the projects IP-2016-06-7686 “Retrieval of Information from Different Optical 3D Remote Sensing Sources for Use in Forest Inventory (3D-FORINVENT)” and IP-2016-06-5621 “Geospatial Monitoring of Green Infrastructure by Means of Terrestrial, Airborne and Satellite Imagery (GEMINI)”. The authors wish to thank Hrvatske vode, Zagreb, Croatia, for providing ALS data.



REFERENCES

  1. COLOMINA I, MOLINA P 2014 Unmanned aerial systems for photogrammetry and remote sensing: A review. ISPRS J Photogramm Remote Sens 92: 79-97. DOI: https://doi.org/10.1016/j.isprsjprs.2014.02.013
  2. TANG L, SHAO G 2015 Drone remote sensing for forestry research and practices. J For Res 26 (4): 791-797. DOI: https://doi.org/10.1007/s11676-015-0088-y
  3. TORRESAN C, BERTON A., CAROTENUTO F, DI GENNARO SF, GIOLI B, MATESE A, MIGLIETTA F, VAGNOLI, C, ZALDEI A, WALLACE L 2017 Forestry applications of UAVs in Europe: a review. Int J Remote Sens 38 (8-10): 2427-2447. DOI: https://doi.org/10.1080/01431161.2016.1252477
  4. NÄSI R, HONKAVAARA E, LYYTIKÄINEN-SAARENMAA P, BLOMQVIST M, LITKEY P, HAKALA T, VILJANEN N, KANTOLA T, TANHUANPÄÄ T, HOLOPAINEN M 2015 Using UAV-based photogrammetry and hyperspectral imaging for mapping bark beetle damage at tree-level. Remote Sens 7 (11): 15467-15493. DOI: https://doi.org/10.3390/rs71115467
  5. SMIGAJ M, GAULTON R, BARR SL, SUÁREZ JC 2015 UAV-borne thermal imaging for forest health monitoring: detection of disease-induced canopy temperature increase. ISPRS - Int Arch Photogramm Remote Sens Spat Inf Sci XL-3/W3: 349-354. DOI: https://doi.org/10.5194/isprsarchives-XL-3-W3-349-2015
  6. DASH JP, WATT MS, PEARSE GD, HEAPHY M, DUNGEY HS 2017 Assessing very high resolution UAV imagery for monitoring forest health during a simulated disease outbreak. ISPRS J Photogramm 131: 1-14. DOI: https://doi.org/10.1016/j.isprsjprs.2017.07.007
  7. PULITI S, ØRKA HO, GOBAKKEN T, NÆSSET E 2015 Inventory of Small Forest Areas Using an Unmanned Aerial System. Remote Sens 7 (8): 9632-9654. DOI: https://doi.org/10.3390/rs70809632
  8. KACHAMBA DJ, ØRKA HO, GOBAKKEN T, EID T, MWASE W 2016 Biomass Estimation Using 3D Data from Unmanned Aerial Vehicle Imagery in a Tropical Woodland. Remote Sens 8 (11): 968. DOI: https://doi.org/10.3390/rs8110968
  9. GETZIN S, NUSKE RS, WIEGAND K 2014 Using Unmanned Aerial Vehicles (UAV) to Quantify Spatial Gap Patterns in Forests. Remote Sens 6 (8): 6988-7004. DOI: https://doi.org/10.3390/rs6086988
  10. HÖHLE J, HÖHLE M 2009 Accuracy assessment of digital elevation models by means of robust statistical methods. ISPRS J Photogramm 64 (5): 398-406. DOI: https://doi.org/10.1016/j.isprsjprs.2009.02.003
  11. GUERRA-HERNÁNDEZ J, GONZÁLEZ-FERREIRO E, SARMENTO A, SILVA J, NUNES A, CORREIA AC, FONTES L, TOMÉ M, DÍAZ-VARELA R 2016 Using high resolution UAV imagery to estimate tree variables in Pinus pinea plantation in Portugal. Forest Syst 25 (2): eSC09. DOI: https://doi.org/10.5424/fs/2016252-08895
  12. TUOMINEN S, BALAZS A, SAARI H, PÖLÖNEN I, SARKEALA J, VIITALA R 2015 Unmanned aerial system imagery and photogrammetric canopy height data in area-based estimation of forest variables. Silva Fenn 49 (5): 1348. DOI: https://doi.org/10.14214/sf.1348
  13. WALLACE L, LUCIEER A, MALENOVSKỲ Z, TURNER D, VOPĚNKA P 2016 Assessment of forest structure using two UAV techniques: A comparison of airborne laser scanning and structure from motion (SfM) point clouds. Forests 7 (3): 62. DOI: https://doi.org/10.3390/f7030062
  14. HIRD JN, MONTAGHI A, MCDERMID GJ, KARIYEVA J, MOORMAN BJ, NIELSEN SE, MCINTOSH A 2017 Use of unmanned aerial vehicles for monitoring recovery of forest vegetation on petroleum well sites. Remote Sens 9 (5): 413. DOI: https://doi.org/10.3390/rs9050413
  15. MLAMBO R, WOODHOUSE IH, GERARD F, ANDERSON K 2017 Structure from motion (SfM) photogrammetry with drone data: A low cost method for monitoring greenhouse gas emissions from forests in developing countries. Forests 8 (3): 68. DOI: https://doi.org/10.3390/f8030068
  16. NEX F, REMONDINO F 2014 UAV for 3D mapping applications: A review. Appl Geomatics 6 (1): 1-15. DOI: https://doi.org/10.1007/s12518-013-0120-x
  17. GAŠPAROVIĆ M, GAJSKI D 2016 The algorithm for the precise elimination of lens distortion influence on digital cameras. Geod List 70 (1): 25-38. URL: http://hrcak.srce.hr/file/231227 (1 August 2017)
  18. GAŠPAROVIĆ M, GAJSKI D 2016 Two-step camera calibration method developed for micro UAV’S. ISPRS - Int Arch Photogramm Remote Sens Spat Inf Sci XLI-B1: 829-833. DOI: https://doi.org/10.5194/isprsarchives-XLI-B1-829-2016
  19. CHIKHRADZE N 2015 Close Range Photogrammetry in the Survey of the Coastal Area Geoecological Conditions (on the Example of Portugal). Earth Sci 4 (5): 35. DOI: https://doi.org/10.11648/j.earth.s.2015040501.17
  20. VANDER JAGT B, LUCIEER A, WALLACE L, TURNER D, DURAND M 2015 Snow Depth Retrieval with UAS Using Photogrammetric Techniques. Geosci 5 (3): 264-285. DOI: https://doi.org/10.3390/geosciences5030264
  21. FAZELI H, SAMADZADEGAN F, DADRASJAVAN F 2016 Evaluating the potential of RTK-UAV for automatic point cloud generation in 3D rapid mapping. ISPRS - Int Arch Photogramm Remote Sens Spat Inf Sci XLI-B6: 221-226. DOI: https://doi.org/10.5194/isprsarchives-XLI-B6-221-2016
  22. GAŠPAROVIĆ M, JURJEVIĆ L 2017 Gimbal Influence on the Stability of Exterior Orientation Parameters of UAV Acquired Images. Sensors 17 (2): 401. DOI: https://doi.org/10.3390/s17020401
  23. BALTSAVIAS E, GRUEN A, EISENBEISS H, ZHANG L, WASERL T 2008 High-quality image matching and automated generation of 3D tree models. Int J Remote Sens 29 (5): 1243–1259. DOI: https://doi.org/10.1080/01431160701736513
  24. ST‐ONGE B, VÉGA C, FOURNIER RA, HU Y 2008 Mapping canopy height using a combination of digital stereo‐photogrammetry and lidar. Int J Remote Sens 29 (11): 3343–3364. DOI: https://doi.org/10.1080/01431160701469040
  25. BALENOVIĆ I, MARJANOVIĆ H, VULETIĆ D, PALADINIĆ E, OSTROGOVIĆ SEVER MZ, INDIR K 2016 Quality assessment of high density digital surface model over different land cover classes. Period Biol 117 (4): 1-12. DOI: https://doi.org/10.18054/pb.2015.117.4.3452
  26. REHAK M 2017 Integrated Sensor Orientation on Micro Aerial Vehicles. Phd Thesis, Swiss Federal Institute of Technology Lausanne (EPFL), Lausanne, Switzerland, 227 p. DOI: https://doi.org/10.5075/epfl-thesis-7530
  27. NI W, LIU J, ZHANG Z, SUN G, YANG A 2015 Evaluation of UAV-based forest inventory system compared with LiDAR data. IGARSS 2015: 3874-3877. DOI: https://doi.org/10.1109/IGARSS.2015.7326670
  28. VEPAKOMMA U, CORMIER D 2015 Inventorying the forest: Laser scanning vs close range photogrammetry on a UAV. In 14th conference in a series focusing on Lidar Applications for assessing and Managing Forest Ecosystems-SilviLaser, France, La Grande Motte, 28-30 September 2015. The French Society of Photogrammetry and Remote Sensing (SFPT), Paris, France, pp 68-70. URL: https://silvilaser2015.teledetection.fr/files/Proceedings_Silvilaser_22_09_2015_2.pdf (21 August 2017)
  29. THIEL C, SCHMULLIUS C 2017 Comparison of UAV photograph-based and airborne lidar-based point clouds over forest from a forestry application perspective. Int J Remote Sens 38 (8-10): 2411-2426. DOI: https://doi.org/10.1080/01431161.2016.1225181
  30. SHAPIRO SS, WILK MB 1965 An analysis of variance test for normality (complete samples). Biometrika 52 (3-4): 591-611. DOI: https://doi.org/10.1093/biomet/52.3-4.591
  31. SHAPIRO SS, WILK MB, CHEN HJ 1968 A comparative study of various tests for normality. J Am Stat Assoc 63 (324): 1343-1372. DOI: https://doi.org/10.1080/01621459.1968.10480932
  32. HILL T, LEWICKI P 2007 STATISTICS: Methods and Applications. StatSoft, Inc., Tulsa, OK, USA, 800 p
  33. R CORE TEAM 2017 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL: https://www.R-project.org/ (3 September 2017)
  34. BLUE MARBLE GEOGRAPHICS 2017 Global Mapper 19. URL: http://www.bluemarblegeo.com/products/global-mapper.php (4 September 2017)
  35. QGIS DEVELOPMENT TEAM 2017 QGIS Geographic Information System. Open Source Geospatial Foundation. URL: http://qgis.osgeo.org (14 September 2017)
  36. HÖHLE J, PEDERSEN CØ, BAYER T, FREDERIKSEN P 2010 The photogrammetric derivation of digital terrain models in built-up areas. Photogrammetric Journal of Finland 22 (1): 33-45
  37. NURMINEN K, KARJALAINEN M, YU X, HYYPPÄ J, HONKAVAARA E 2013 Performance of dense digital surface models based on image matching in the estimation of plot-level forest variables. ISPRS J Photogramm 83: 104-115. DOI: https://doi.org/10.1016/j.isprsjprs.2013.06.005
  38. HOBI ML, GINZLER C 2012 Accuracy Assessment of Digital Surface Models Based on WorldView-2 and ADS80 Stereo Remote Sensing Data. Sensors 12 (5): 6347-6368. DOI: https://doi.org/10.3390/s120506347
  39. WHITE JC, STEPPER C, TOMPALSKI P, COOPS NC, WULDER MA 2015 Comparing ALS and Image-Based Point Cloud Metrics and Modelled Forest Inventory Attributes in a Complex Coastal Forest Environment. Forests 6 (10): 3704-3732. DOI: https://doi.org/10.3390/f6103704 
  40. VALBUENA R, MAURO F, ARJONILLA FJ, MANZANERA JA 2011 Comparing airborne laser scanning-imagery fusion methods based on geometric accuracy in forested areas. Remote Sens Environ 115 (8): 1942-1954. DOI: https://doi.org/10.1016/j.rse.2011.03.017
  41. SIMIC MILAS A, AREND K, MAYER C, SIMONSON MA, MACKEY S 2017 Different Colors of Shadows: Classification of UAV images. Int J Remote Sens 38 (8-10): 3084-3100. DOI: https://doi.org/10.1080/01431161.2016.1274449

© 2017 by the Croatian Forest Research Institute. This is an Open Access paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0).

- Download PDF-

SEEFOR 8 (2): 147-149
Article ID: 167
DOI: https://doi.org/10.15177/seefor.17-14

Preliminary communication

 

First Record of Biocontrol Agent Torymus sinensis (Hymenoptera; Torymidae) in Bosnia and Herzegovina


Dinka Matošević1*, Osman Mujezinović2, Mirza Dautbašić2


(1) Croatian Forest Research Institute, Department for Forest Protection and Game Manage-ment, Cvjetno naselje 41, HR-10450 Jastrebarsko, Croatia;
(2) University of Sarajevo, Faculty of Forestry, Chair for Forest and Urban Greening Protection, Wildlife and Hunting, Zagrebačka 20, BA-71000 Sarajevo, Bosnia and Herzegovina

* Correspondence: e-mail:

Citation: MATOŠEVIĆ D, MUJEZINOVIĆ O, DAUTBAŠIĆ M 2017 First Record of Biocontrol Agent Torymus sinensis (Hymenoptera; Torymidae) in Bosnia and Herzegovina. South-east Eur for 8 (2): 147-149. DOI: https://doi.org/10.15177/seefor.17-14

Received: 6 Oct 2017; Accepted: 12 Nov 2017; Published online: 24 Nov 2017


Cited by:     Crossref  (0)      Google Scholar


Abstract

Background and Purpose: Dryocosmus kuriphilus is an invasive insect species of sweet chestnut (Castanea spp.) originating from China, and the only effective control measure against this pest is classical biological control with introduced parasitoid Torymus sinensis. This parasitoid has been widely released in many European countries, but it also has the ability to rapidly spread naturally. No official releases have been done in Bosnia and Herzegovina.
Material and Methods: D. kuriphilus galls were collected in July 2017 on 6 localities in forest district Unsko (Una Sana canton) in Bosnia and Herzegovina. Presence and parasitism rates of T. sinensis were recorded in the entomological laboratory, Croatian Forest Research Institute. T. sinensis larvae were identified morphologically and by being compared with the voucher specimens.
Results and Discussion: Torymus sinensis larvae were positively identified in the examined D. kuriphilus galls from all localities in Bosnia and Herzegovina. Parasitism rates ranged from 44.83 to 74%. Occurrence and high parasitism rates in Bosnia and Herzegovina observed in this study are not results of biocontrol releases of T. sinensis, but can be attributed to natural spread from Croatia. High parasitism rates observed in this study can indicate that the parasitoid was present in Bosnia and Herzegovina in 2016.
Conclusions: This study presents the first record of Torymus sinensis in Bosnia and Herzegovina. We predict that the parasitoid will continue its spread over Bosnia and Herzegovina in sweet chestnut forests and orchards and that it will act as effective biological control agent against D. kuriphilus.

Keywords: parasitoid, invasive species, Dryocosmus kuriphilus, natural spread, classical biological control, parasitism rate 



INTRODUCTION

Dryocosmus kuriphilus Yasumatsu (Hymenoptera; Cynipidae) is an invasive insect species, originating from China, which has spread in sweet chestnut (Castanea spp.) forests and orchards around the word [1]. In Europe, it has been first introduced to Italy [2] and from there it spread to the majority of European countries [1]. In Bosnia and Herzegovina D. kuriphilus was first recorded in 2015 in Una Sana canton [3]. D. kuriphilus is regarded as a serious threat to chestnuts, especially to fruit production [4] and crown leaf area loss [5]. Parasitoid Torymus sinensis Kamijo [Hymenoptera; Torymidae] has successfully been used as a classical biological control agent against D. kuriphilus and it has been released in biocontrol campaigns in Japan, the USA, Italy, France, Slovenia, Croatia and Hungary [6, 7, 8, 9, 10, 11]. Torymus sinensis is native to China, phenologically well synchronised with D. kuriphilus, it is highly specific, and lowers the outbreak levels of its host [9, 12, 13]. This parasitoid has shown high dispersal ability by being able to cover more than 70 km in only a few days aided by wind [13]. Croatia has done extensive biocontrol releases of T. sinensis since 2014, and apart from releases, the parasitoid has also rapidly spread from Italy over Slovenia to Croatia and has built a viable population with parasitism rates up to 90% [11, 14]. Based on this experience, we have expected T. sinensis to spread towards Bosnia and Herzegovina. Bosnia and Herzegovina has so far done no official releases of T. sinensis on its territory.

The aim of this paper is to report first record of biocontrol agent T. sinensis and its parasitism rates in sweet chestnut (Castanea sativa Mill.) forests of Bosnia and Herzegovina.

 

MATERIALS AND METHODS

D. kuriphilus galls were collected in July 2017 on 6 localities in forest district Unsko (Una Sana canton) in Bosnia and Herzegovina (Table 1). The galls were collected from randomly selected sweet chestnut trees from a height of 1.5-2.5 m. From each locality a sample of 100 galls was taken. Each gall from the sample was sliced open and examined under a binocular microscope Olympus SZX7 in entomological laboratory, Croatian Forest Research Institute. In dissected galls, larval chambers, the number of T. sinensis larvae, D. kuriphilus larvae and pupae (if present), and other parasitoid larvae were counted and parasitism rates were calculated: PR = (the number of T. sinensis specimens/the number of D. kuriphilus larval chambers) ×100 (%). T. sinensis larvae were identified morphologically [11, 15] and by being compared with the voucher specimens deposited at the Department for Forest Protection, Croatian Forest Research Institute. The larvae were stored in absolute ethanol at -20°C in entomological laboratory, Croatian Forest Research Institute, for further analyses.

 

RESULTS AND DISCUSSION

Torymus sinensis larvae were positively identified in examined D. kuriphilus galls from all six localities. Parasitism rates ranged from 44.83 to 74% (Table 1).

 

TABLE 1. Localities, coordinates and parasitism rates of Torymus sinensis for the samples taken in Una Sana canton, Bosnia and Herzegovina, in 2017.

 

The results of our study show presence of T. sinensis in Bosnia and Herzegovina. Occurrence and high parasitism rates in Bosnia and Herzegovina observed in this study (Table 1) are not results of biocontrol releases of T. sinensis, but can be attributed to natural spread [13, 14]. This biocontrol agent spread naturally from Croatia over interconnected sweet chestnut forests and wooded chestnut patches bordering Croatia and Bosnia and Herzegovina in Una Sana canton. It has already been documented that T. sinensis is rapidly spreading naturally eastwards from Italy all over Croatia [14]. This rapid natural spread was additionally assisted by releases from laboratory rearing in Croatia in the area near the border with Bosnia and Herzegovina in 2016 and 2017 [14]. High parasitism rates observed in this study can indicate that the parasitoid was present in Bosnia and Herzegovina in 2016, but was not sampled and identified. We predict that the parasitoid will continue its spread over Bosnia and Herzegovina in sweet chestnut forests and orchards and that it will act as effective biological control agent against D. kuriphilus, lowering its population and damages in sweet chestnut forests.

 

Acknowledgments

Dinka Matošević would like to thank Blaženka Ercegovac and Ivana Mihaljević for laboratory work.


REFERENCES

  1. CABI 2017 Dryocosmus kuriphilus. In: Invasive species compendium. URL: https://www.cabi.org/isc/datasheet/20005 (10 September 2017).
  2. BRUSSINO G, BOSIO G, BAUDINO M, GIORDANO R, RAMELLO F, MELIKA G 2002 Dangerous exotic insect for the European chestnut. Informatore Agrario 58 (37): 59-61
  3. DELALIC Z 2016 Prvi nalaz karantinske štetočine kestenove ose šiškarice (Drycosmus kuriphilus) u Bosni i Hercegovini. Biljni Lekar 44 (1): 58-65
  4. BATTISTI A, BEVEGNU I, COLOMBARI F, HAACK RA 2013 Invasion by the chestnut gall wasp in Italy causes significant yield loss in Castanea sativa nut production. Agr For Entomol 16 (1): 75-79. DOI: https://doi.org/10.1111/afe.12036
  5. GEHRING E, BELLOSI B, QUACCHIA A, CONEDERA M 2017 Assessing the impact of Dryocosmus kuriphilus on the chestnut tree: branch architecture matters. J Pest Sci. DOI: https://doi.org/10.1007/s10340-017-0857-9
  6. MORIYA S, SHIGA M, ADACHI I 2003 Classical biological control of the chestnut gall wasp in Japan. In: van Driesche RG (ed) Proceedings of the 1st international symposium on biological control of arthropods, USDA Forest Service, Washington, pp 407-415
  7. COOPER WR, RIESKE LK 2011 A native and an introduced parasitoid utilize an exotic gall-maker host. BioControl 56 (5): 725-734. DOI: https://doi.org/10.1007/s10526-011-9350-1
  8. BOROWIEC N, THAON M, BRANCACCIO L, WAROT S, VERCKEN E, FAUVERGUE X, RIS N, MALAUSA JC 2014 Classical biological control against the chestnut gall wasp Dryocosmus kuriphilus (Hymenoptera, Cynipidae) in France. Plant Protection Quarterly 29 (1): 7-10
  9. MATOŠEVIĆ D, QUACCHIA A, KRISTON É, MELIKA G 2014 Biological Control of the Invasive Dryocosmus kuriphilus (Hymenoptera: Cynipidae) - an Overview and the First Trials in Croatia. South-east Eur for 5 (1): 3-12. DOI: https://doi.org/10.15177/seefor.14-05
  10. FERRACINI C, GONELLA E, FERRARI E, SALADINI MA, PICCIAU L, TOTA F, PONTINI M, ALMA A 2015 Novel insight in the life cycle of Torymus sinensis, biocontrol agent of the chestnut gall wasp. BioControl 60 (2): 169-177. DOI: https://doi.org/10.1007/s10526-014-9633-4
  11. MATOŠEVIĆ D, LACKOVIĆ N, MELIKA G, KOS K, FRANIĆ I, KRISTON E, BOZSO M, SELJAK G, ROT M 2015 Biological control of invasive Dryocosmus kuriphilus with introduced parasitoid Torymus sinensis in Croatia, Slovenia and Hungary. Periodicum Biologorum 117 (4): 471-477
  12. QUACCHIA A, MORIYA S, BOSIO G, SCAPIN I, ALMA A 2008 Rearing, release and settlement prospectin Italy of Torymus sinensis, the biological control agent of the chestnut gall wasp Dryocosmus kuriphilus. BioControl 53 (6), 829-839. DOI: https://doi.org/10.1007/s10526-007-9139-4
  13. COLOMBARI F, BATTISTI A 2016 Spread of the introduced biocontrol agent Torymus sinensis in north-eastern Italy: dispersal through active flight or assisted by wind? BioControl 61 (2): 121-139. DOI: https://doi.org/10.1007/s10526-015-9712-1
  14. MATOŠEVIĆ D, LACKOVIĆ N, KOS K, KRISTON E, MELIKA G, ROT M, PERNEK M 2017 Success of classical biocontrol agent Torymus sinensis within its expanding range in Europe. J Appl Entomol 141 (9): 758-767. DOI: https://doi.org/10.1111/jen.12388
  15. GÓMEZ JF, NIEVES-ALDREY JL, NIEVES MH 2008 Comparative morphology, biology and phylogeny of terminal-instar larvae of the European species of Toryminae (Hym., Chalcidoidea, Torymidae) parasitoids of gall wasps (Hym. Cynipidae). Zool J Linn Soc-Lon 154 (4): 676-721. DOI: https://doi.org/10.1111/j.1096-3642.2008.00423.x


© 2017 by the Croatian Forest Research Institute. This is an Open Access paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0).