Geostatistical analysis of spatial variation in forest ecosystems

Eurasscience Journals Eurasian Journal of Forest Science (2018) 6(1): 9-22 Geostatistics in characterizing spatial variability of forest ecosystems Gülay Karahan1*, Sabit Erşahin2 1) Cankırı Karatekin University, Faculty of Forestry, Department of Landscape Architecture, 18200, Cankırı,Turkey 2) Cankırı Karatekin University, Faculty of Forestry, Department of Forest Engineering, 18200, Cankırı,Turkey *corresponding author: Abstract Forests are spatially variable due to multiple interactions among state (vegetation, species distribution, understory cover, soil, and topography) and forcing variables (climate and human) variables. In general, the spatial structure is resulted as combined effect of these external and internal variables. Geostatistical methods can aid characterizing the spatial structure of forest ecosystems. The shape and parameters (nugget, sill, range) of semivariograms provide important information on the characteristics of spatial structure. In addition, the geostatistical interpolation methods (e.g. kriging) are effective tools for constructing surface maps of variable of interest. Thus, the geostatistical methods have been used increasingly for characterizing forest spatial structure across different spatial scales for last 30 years. In this literature study, sources of spatial variability of forest ecosystems are explained and results of several geostatistical studies are discussed. Keywords: Nugget, Range, Sill, Spatial interpolation, Spatial structure, Özet Ormanlar zorlayıcı (dışsal) ve etkilenen (durum) değişkenleri arasındaki çoklu etkileşimler nedeniyle uzaysal değişkenlik gösterirler. Genel olarak, uzaysal değişkenlik bu değişkenlerin ortak etkisinin bir sonucu olarak ortaya çıkmaktadır. Joeistatiksel yöntemler uzaysal yapının karakterize edilmesine yardımcı olabilmektedir. Semivaryogramın şekli ve parametreleri (nugget, sill, range) uzaysal yapı hakkında önemli bilgiler sağlar. Ayrıca, jeoistiksel enterpolasyon yöntemleri (örneğin, krigleme) ilgili değişkenin yüzey haritalarının çıkarılmasında oldukça kullanışlı araçlardır. Dolayısıyla, jeoistatistiksel yöntemler son 30 yılda ormanların uzaysal değişkenliklerinin karakterize edilmesinde artan bir şekilde kullanılmaktadır. Bu literatür çalışmasında, ormanların uzaysal değişkenliğinin başlıca kaynakları verildikten sonra, bu kaynakların bir fonksiyonu olarak ortaya çıkan uzaysal değişkenliğin karakterize edilmesinde yapılmış bazı jeoistatiksel çalışmaların sonuçları tartışılmıştır. Anahtar kelimeler: İklim, Orman ekosistemleri, Nugget, Sill, Range, Uzaysal yapı 9 Eurasian Journal of Forest Science 6(1):9-22 (2018) Introduction Forest ecosystems vary in time and space. Spatially continuous data are important in all ecosystems including forests for decision-making. Therefore, analysis of spatial variability of forest ecosystems is needed for its thorough understanding. In addition, understanding spatial variation of forests improves our understanding of ecosystem-level processes. According to Pelissari et al. (2017), deficiency of ecological information needs to new techniques for analyzing spatial variations in forests, one of them, geostatistics is a technique for modeling and mapping. Geostatistics was generally applied in forest research (Akhavan et al., 2010; Fox et al., 2007; Nanos et al., 2004; Palmer et al., 2010; Pelissari et al., 2014; Sales et al., 2007 ). The geostatistical methods are robust because the area of influence can be adjusted according to the case study needs (Torres et al., 2017). Predicting values of a variable in unsampled points allows to generate spatially continuous data (Li and Heap 2008). Goal of geostatistics is to examine the spatial structure of the target variable and predict its values at unsampled locations. Therefore, geostatistics is an important technique that can be used to characterize spatial or temporal phenomena (Zhang, 2011). Geostatistics includes ways for analyzing the autocorrelation in spatial data. An important property of geostatistics is the semivariance, which measures spatial continuity. Use of the semivariograms needs the data supplies the real hypothesis for regional variable (Journel and Huijbregts, 1978). There have been number of studies carried out on forest ecosystems. Most of these studies were focused on carbon storage, forest biomass, growth rate and variability of trees, and forest soil quality etc. When compared with the others, geostatistics gives a powerful way to make easy of the spatial variation and interpolation quantification. In this study, geostatistical analysis of forest spatial variability as related to topography, land use, soils, and climate are mentioned and results of several studies are discussed as well. Geostatistical measures of spatial variability in forest ecosystems Field measurements are basic requirement in collecting information on forests. But, these measurements can be cost, time consuming and impractical in large areas (Zawadzki et al. 2005). According to Clark (1979), conventional statistics cannot completely explain the spatial variations. Therefore, geostatistical methods ensure a probabilistic structure for understanding the characteristics of the spatial distribution of forest variables (Zhang, 2011). According to Isaak and Srivastava (1989) and Goovaerts (1997), geostatistics was improved to analyze variables, which are distributed continually in space, called "regionalized variables". The aim of geostatistics is the prediction of values of a target attribute at unsampled locations. Key steps for defining and estimating are 1) modeling of the spatial variability of data of the property by fitting of models to the experimental semivariogram, and 2) using the data with parameters of theoretical semivariogram to interpolate the target attribute in the study area (Goovaerts, 1998). Steps of analyzing spatial pattern 1- The histograms of the data (pH in this example) are plotted and summary statistics are computed (Fig. 1). However, by this way, critical information such as spatial location of pH measurements cannot be gained (Goovaerts 1998). 2- Each values along the transect does not distribute completely random. Because close observations tend to be like. For example, h-scattergram of the pH values can be showed by plotting with observations separated by a distance of 1-m (Figure 2) (Goovaerts 1998). 10 Geostatistics in characterizing spatial variability - Karahan and Erşahin - 6(1):9-22 (2018) Fig. 1. Histograms of soil pH values measured in a forest plot. Fig.2. Scattergram of the soil pH values 3- The image of the graph shows correlations of pH values. These correlations evaluate with the linear correlation coefficient. By plotting of the estimated correlation coefficients, experimental correlogram is obtained (Fig.3). Fig. 3. Correlogram and semivariance of soil pH values measured in forest (Goovaerts 1998). 4- Spatial patterns are described with differences in data pairs. F (...truncated)