Printer Friendly

Predicting landscape structural metrics using ASTER satellite data/Krastovaizdzio strukturiniu metriku nustatymas remiantis ASTER palydoviniais duomenimis.

1. Introduction

An important challenge for forest managers is how to maintain biological diversity and to integrate it into existing forest management plan (Baskent et al. 2008; Ozkan, Mert 2011). It is widely agreed in the literature that biological diversity is best addressed at the landscape level when managing forests (Baskent, Jordan 1995, 1996; Schindler et al. 2008). One straightforward way to maintain biological diversity that is recommended by many ecologists is to maintain landscape diversity as a structurally diverse landscape provides foraging, breeding, and hiding habitat for numerous species (Harris, Smith 1978; Smith 2000; Fortin, Melles 2009). Therefore, the spatial structure of forest landscape should be controlled and enhanced by proper management systems that mimic natural spatial patterns (Baskent 1997, 1999; Kucas 2010).

The spatial structure of forest landscape describes the composition and configuration of landscape patches. It changes over time by both management activities including regeneration, reforestation, rehabilitation and disturbances such as fire, insect attack, windstorm and urbanization, etc. Monitoring the changes in landscape structure by definite time intervals and taking necessary silvicultural precautions is vital for sustainable management of forest ecosystems. Existing digital forest maps provide a good source of information to characterize spatial structure. Based on these maps, landscape spatial structure can be quantified by using landscape structural metrics (McGarigal et al. 2002). Using Geographical Information System (GIS) maps, landscape spatial structure can be mapped using either a static (geographic stratification) or a dynamic approach (moving window). In the static approach, the reference units are pre-defined and then landscape metrics are calculated for each reference unit. On the contrary, in the dynamic approach, the metrics are calculated using a window moving across the landscape. In this approach, a structural index value is assigned to the centre of each window and then the index values are transferred to an output raster image. In the next step, the index value of each cell of the output image is determined by means of the surrounding index values. Bilinear interpolation algorithm can be used to calculate index value for each image cell (Eiden et al. 2000). Therefore, a landscape structure map can be obtained using these approaches, which is the basis for future monitoring tasks. Such diversity maps will attract wildlife managers' attention and may help forest managers to integrate biodiversity into existing management plans.

Although GIS-ready forest maps are a convenient resource for mapping and monitoring landscape structure (Bauza 2007), they are often out of date for extensive areas. Wide-area satellite remote sensing appears to be a useful alternative to forest maps. Remotely sensed data has been widely used in biodiversity studies including prediction of species richness (Rocchini et al. 2009; Rocchini 2009), canopy heterogeneity (Hyde et al. 2006; Goetz et al. 2007), tree size diversity (Ozdemir et al. 2008; Ozdemir, Karnieli 2011), and characterization of forest spatial structure (Egbert et al. 2002; Kadiogullari, Baskent 2008). Digital satellite data may be used to estimate landscape metrics in case there is a relation between image features and real landscape metrics. If a strong relation exists, the structural diversity can be mapped inexpensively using satellite data. Accordingly, the potential of satellite data should be investigated in order to predict and map landscape pattern characteristics.

Spatial scale is one of the most important issues in landscape ecology. Different organisms have differential response to environmental gradients at different scales. Spatial scale could be both 'grain' (spatial resolution of the image or the mapping unit) and 'extent' (window size and/or extent of the study area) (Gustafson 1998). When landscape extent increases, more patch types can exist, resulting in higher landscape diversity (Turner et al. 1989; Peters, Goslee 2000). For that reason, the landscape extent should be taken into consideration in estimating landscape structural metrics by employing remotely sensed data.

The main objective of this study was, therefore, to explore the relations between landscape metrics calculated on the basis of forest stand polygons and the features extracted from ASTER imagery with 15 m resolution. In addition, the study aimed to determine the best landscape extent to predict landscape diversity using ASTER data.

2. Materials and methods

In this study, firstly, landscape pattern metrics were calculated for different landscape extents based on manually delineated stand polygons from aerial photos. Secondly, landscape metrics were calculated for the same extents using the polygons automatically generalized from the ASTER imagery by multi-resolution segmentation method. Thirdly, texture image features were derived from the landscape extents. Lastly, the relationships between landscape metrics and image derived features were determined using the stepwise multiple linear regression analysis.

2.1. Data preparation

2.1.1. The GIS-ready data

The five digital forest maps of the Isparta Forest Regional Directorate prepared for the purpose of forest management in 2006 and 2007 were obtained. Forest stand delineation was performed based primarily on aerial photo (1/15000 scaled) interpretation in these maps. It was decided that the forest stand maps prepared for timber management purpose were too detailed to analyze spatial landscape structure due to the similar habitat characteristics of several stand classes. Therefore, such stand classes were combined in the GIS database. Moreover, there were many adjacent polygons that belong to the same stand class in these maps because a forest stand can normally be divided into sub-compartments within a main compartment in the mapping system used in Turkey. The adjacent polygons with the same attribute (patch class) were merged to one polygon. Also, the roads or firebreaks defined as lines on the map were converted into a strip with a width of 10 m and then, they were attributed as forest openings in the database. The last modification was made for the clear-felled areas practiced after the map preparation date and before the image acquisition date. These areas were easily defined and delineated from the ASTER imagery due to their remarkable spectral response.

2.1.2. The image data

The image features were derived from the three bands with 15 m spatial resolution of ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) satellite data with an image acquisition date of 5 May 2007. The spectral ranges of the three bands were 0.52-0.60 [micro]m (green), 0.63-0.69 [micro]m (red) and 0.76-0.86 [micro]m (near infrared). The image was clipped based on the boundaries of the study area. For a precise match, the clipped image was geo-referenced by means of well-defined image features such as the intersections of streams and roads on both the forest map and the image. The nearest neighbor re-sampling method was used to preserve the original pixel values. The root mean squared error was 10 m, which is less than one pixel (15 m).

2.2. Calculation of response landscape structural metrics

The landscape metrics were calculated on the basis of square-shaped windows using the forest map polygons. For this purpose, study area has been divided into square units with the size of 225 ha (1500 by 1500 m) and then, 50 sample units were randomly selected for evaluation (Fig. 1). Taking the top-left corner of square-shaped units as a reference point, the three extents were also formed, which are 1200 x 1200 m (144 ha), 900 x 900 m (81 ha) and 600 x 600 m (36 ha) (Fig. 2). Thus, we employed four landscape extents (sizes) to test which one is best to model landscape pattern metrics using the features extracted from ASTER data.


The 36 ha area was determined as the minimum landscape extent because the areas smaller than this were not sufficient for calculating some landscape metrics (e.g. IJI and SEI). The 225 ha area was determined as the maximum landscape extent because the variation in the landscape metrics among sample units was decreased when the landscape extent was larger than 225 ha, which may reduce the correlation between landscape metrics and remotely sensed variables.

For each of the four landscape extents, landscape patch polygons were intersected by the 50 square-shaped units. In the next step, each square-shaped unit was done independent by assigning them a region number. Then, the units in vector format were converted into a raster image. Lastly, the landscape metrics were calculated for each sample unit independently using the ArcView 3.1. Patch Analyst extension software.


Many landscape metrics have been developed to quantify spatial landscape characteristics. This study focused on the three groups of statistics provided by the Patch Analyst, which are; i) patch size and density, ii) edge metrics, iii) diversity and interspersion metrics. The six landscape metrics were selected from these categories, including Number of Patches (NUMP), Edge density (ED), Interspersion Juxtaposition Index (IJI), Shannon's Diversity Index (SDI), Shannon's Evenness Index (SEI) and Patch richness (PR). The formulas and their units of the landscape metrics were given below (McGarigal et al. 2002):

NUMP = N; (1)

ED = E/A (10000); (2)

IJI = [[-[m.summation over (i=1)][m.summation over (k=i+1)][([e.sub.ik]/E) x ln ([e.sub.ik]/E)]]/[ln(0.5[m(m-1)])]] (3)

SDI = -[m.summation over (i=1)] ([P.sub.i] ln [P.sub.i]); (4)

SEI = [[-[m.summation over (i=1)]([P.sub.i] ln [P.sub.i])]/ln m]; (5)

PR = m, (6)

where, N is the total number of patches in the landscape, E is the total length of edge in the landscape, A is total landscape area ([m.sup.2]), [e.sub.ik] is the total length of edge in the landscape between patch classes i and k, m is the number of patch classes present in the landscape, [P.sub.i] is the proportion of the area occupied by patch class i.

2.3. Extraction of remote sensing variables

2.3.1. Multi-resolution segmentation polygons

Image segmentation algorithms have been widely used in forestry applications such as in forest mapping (Kim et al. 2011; Chen et al. 2012; Vega, Durrieu 2011; Add-ink et al. 2012), in quantifying landscape heterogeneity (e.g. Morgan, Gergel 2010) and in exploring forest structural complexity (e.g. Lamonaca et al. 2008). The multi-resolution segmentation algorithm provided by Definiens software is a sophisticated method grouping image pixels according to a rule set, including scale parameter and homogeneity criteria. The scale parameter is an abstract term which determines the maximum allowed heterogeneity of the resulting image objects. As the scale parameter is increased, the homogeneity of segments decreases and the standard deviation within the resulting image objects increases. The segmentation process is also guided by defining the homogeneity criteria according to the color shape, compactness and smoothness of the resulting pixel groups (Blaschke 2010). This operation generates ecologically meaningful image objects. The resulting objects should be visually assessed for correct feature delineation since there is no statistical method for evaluating the accuracy of the segmentation process (Baatz et al. 2001). Therefore, several experiments were undertaken toward high-quality segmentation by modifying the scale parameters and homogeneity criteria to capture real landscape features. The resulting image objects were compared to the manually-delineated polygons of the forest maps. As a result of the visual evaluation process, the best segmentation was obtained using the scale parameter of 25 and homogeneity combination of "0.8 color - 0.2 shape" and "0.5 compactness - 0.5 smoothness".

The segmented image object polygons, and their mean brightness values, were exported as a vector file. The segment-based landscape metrics was calculated with regard to the four extents using this vector file (Fig. 3). It should be noted that, the mean brightness values of image segments were used in the evaluation instead of patch classes in the forest map. Consequently, six landscape metrics as explanatory variables were Segment-Based Number of Patches (SB-NUMP), Segment-Based Edge Density (SB-ED), Segment-Based Interspersion Juxtaposition Index (SB-IJI), Segment-Based Shannon's Diversity Index (SB-SDI), Segment-Based Shannon's Evenness Index (SB-SEI) and Segment-Based Patch richness (SB-PR). These explanatory variables were called the "A group" variables.


2.3.2. Image texture parameters

Texture based classification of satellite imagery is a new opportunity and challenge for improving classification accuracy (e.g. Ota et al. 2011; Chang et al. 2011). The texture features also have potential for monitoring insect infestation (e.g. Dye et al. 2008) and evaluating parameters related to biodiversity (e.g. St-Louis et al. 2006; Ozdemir et al. 2008).

Texture features were extracted from the simple chessboard segmentation approach in order to assess the same windows used in the calculation of the landscape metrics. This method simply splits an image into square image objects. According to a given object size, the image objects are generated starting from the top-left corner pixel of an image in this approach provided by Definiens (Baatz et al. 2001). In this case, when using the entire image, we were not able to segment the fixed areas (1500 x 1500 m = 100 x 100 pixels; 1200 x 1200 m = 80 x 80 pixels; 900 x 900 m = 60 x 60 pixels and 600 x 600 m = 40 x 40 pixels) generated for calculating the response landscape metrics. In order to solve this computational limitation, the image was broken up into subsets according to the top-left corners of the 50 sample windows as a reference point in order to run the texture analyses. Consequently, 50 independent windows with an area of 1800 x 1800 m (bigger than the largest landscape extent, 1500 x 1500 m) were segmented by chessboard segmentation approach (Fig. 4). As a result, we generated same extents used in calculating the landscape metrics and the texture parameters were, therefore, precisely matched to the landscape metrics determined for the four landscape extents in the forest map.


The texture parameters of the objects were calculated based on the grey level co-occurrence matrix (GLCM) (Harralick et al. 1973) which is a tabulation of how often different combinations of pixel grey levels occur in a given direction in an image object. Several statistical measures can be derived from the GLCM. These are grouped into three main categories; i) contrast, ii) orderliness and iii) descriptive statistics. Hall-Beyer (2007) recommended using one texture parameter from each category in an examination. Therefore, three texture measures not well correlated with one another, including GLCM Homogeneity (GHO) from the contrast category, GLCM Entropy (GEN) from the orderliness category and GLCM Standard Deviation (GSD) from the descriptive statistics category were examined in this study. Three GLCM parameters were calculated for each of the 50 plots based on the three ASTER bands separately. Consequently, nine texture variables (GHOB1, GHOB2, GHOB3, GENB1, GENB2, GENB3, GSDB1, GSDB2, and GSDB3) were generated for the statistical evaluation. These texture-based explanatory variables were called the "B group" variables. The formulas and their units are given below (Hall-Beyer 2007):

GLCM - Homogeneity = [N-1.summation over (i,j=0)] [[P.sub.i,j]/[1 + [(i - j).sup.2]]]; (7)

GLCM - Entropy = [N-1.summation over (i,j=0)] [P.sub.i,j](-ln [P.sub.i,j]) (8)

GLCM - S.Deviation = [square root of [N-1.summation over (i,j=0)] [P.sub.i,j](i,j - [[mu].sub.i,j])], (9)

where, i is the row number, J is the column number, [V.sub.ij] is the value in the cell i,j of the matrix, [P.sub.ij] is the normalized value in the cell i,j, N is the number of rows or columns, [[mu].sub.i,j] is GLCM mean.

2.4. Statistical analyses

A stepwise multiple linear regression analysis was carried out to identify the most significant image-derived predictors of landscape metrics for each extent. This method determines the variables that make the most contribution to regression model. The adjusted coefficient of determination (adj.[R.sup.2]) was used in assessing the goodness of fit of models. In a multiple linear regression model, adjusted [R.sup.2] measures the proportion of the variation in the dependent variable accounted for by the explanatory variables. Contrary to [R.sup.2], adjusted R square makes allowance for the degrees of freedom associated with the sums of the squares. In spite of the residual sum of squares decreases or remains the same as new explanatory variables are added, the residual variance does not. Therefore, adjusted [R.sup.2] is generally considered to be a more accurate goodness-of-fit measure than [R.sup.2]. The residuals (differences between observed and predicted values) from the model were analyzed by Residual Standard Deviation ([S.sub.res]). It is assumed in multiple regression that the residuals (experimental errors) are distributed normally (Moore and McCabe 1993). Kolmogov-Simirnov Z (K-S) test was used in order to inspect the distribution of the residual values. If P > 0.05, the residuals of a regression model were accepted as having a normal distribution. SPSS package program (13.0) was used for the statistical analyses.

3. Results and discussion

The multiple linear regression models predicting landscape metrics gave different results depending on both landscape extent and the kind of metric. All of the regression models were statistically significant at an alpha level of 0.05. According to K-S test the residuals of the models were distributed normally. In examining the highest [R.sup.2.sub.adj] and lowest [S.sub.res] values, there was no important difference between A and B group variables in terms of explained variance and standard error of estimate. However, Shannon Index, a widely preferred metric, was better modeled using image texture features. Furthermore, since deciding the optimal parameters (e.g. scale parameter) for multi-resolution segmentation is somewhat subjective, the texture parameters based on grey level co-occurrence matrix may be preferred to model the landscape metrics.

The findings showed that all landscape metrics examined here were associated with the image parameters except IJI and SEI at least for one landscape extent. The models providing best fit were obtained using the sizes of 900 x 900 m and 1200 x 1200 m. More than 50% of variance in the Number of Patches, the Edge density, the Shannon's Diversity Index and the Patch richness was explained by the regression models. In contrast, although some models predicting the Interspersion Juxtaposition Index and the Shannon's Evenness Index were statistically significant (p < 0.01, adj. [R.sup.2] is 0.12 for the IJI; p < 0.01, adj. [R.sup.2] is 0.30 for SEI), they might be insufficient for making an estimation of these metrics. The extent of 144 ha (1200 x 1200 m) seems as an effective size for modeling the landscape metrics including the NUMP, SDI, ED and PR by the features of ASTER bands.

The equations of the best regression models in terms of the explained variance and the residual standard deviation are summarized for the A and B group of variables, in the Table 1. All regression models predicting the landscape metrics except the NUMP contained one independent variable, might indicate the existence of high correlation between some independent variables of the A group. The [R.sup.2.sub.adj] and [S.sub.res] values showed that the response NUMP, ED and PR were correlated with the image segment-based landscape metrics. The response ED, SDI and PR were modeled as a function of the SB-ED, SB-SDI and SB-ED respectively that are calculated by same index formulas from the segmented ASTER imagery. Therefore, it can be concluded that automated multi-resolution segmentation yields the patch classes and patch boundaries (edge lengths) that are close to the forest map polygons. Moreover, the multi-resolution segmentation approach may also give a great opportunity to predict the Contrast-Weighted Edge Density (CWED) index more accurately than ED. The CWED taking into consideration edge contrast as well as edge length quantifies edge from the perspective of its functional significance. Edge contrast is defined by structural (e.g. basal area, tree volume and crown size density) and floristic characteristics between adjacent patches composing that edge. Consequently, in addition to edge length (or edge density predicted well in this study), we expected that the edge contrast can be quantified by brightness or texture features of image segments. In conclusion, the response CWED might be more accurately estimated by the satellite-based CWED than the ED.

The NUMP, ED, SDI and PR can be predicted employing the models as a function of the texture parameters of ASTER bands. The GLCM-Entropy calculated from the near infrared band (GENB3) was the most correlated variable as it entered into these multiple models except SEI as a variable. GLCM-Entropy calculated from the red band (GENB2) was also a good predictor, which was included into the regression models except the model of NUMP. Only one model predicting the NUMP contained the GLCM-Homogeneity derived from the red band (GHOB2) as an explanatory variable. Similarly, the GLCM-Standard Deviation extracted from the red band (GSDB2) was present only in the model estimating SDI.

The results demonstrate that the Entropy measure, the level of spatial disorder of grey levels in GLCM, is superior to the other texture measures (the Homogeneity and Standard Deviation) in estimating the landscape metrics. Based on the findings of this study, it is reasonably to conclude that if a landscape (within the limits studied here) has more patch types, edge or number of patch, the GLCM-Entropy value of that landscape is increasing, which shows the elements of GLCM are equally distributed. Another result of this study is that the near-infrared and red bands yielded the strongest explanatory variables of the landscape metrics except IJI. The superiority of these bands over the green band may be due to their capability to distinguish vegetation communities (Lillesand, Kiefer 1994). Similarly, Levin et al. (2007) found that the normalized difference vegetation index (NDVI) calculated using infrared and red bands is a good predictor for species richness. Furthermore, NDVI is correlated with Vegetation Fractional Coverage (Gu et al. 2009). Also, Rocchini (2007) and Rocchini et al. (2007) reported that near infrared band is better than visible bands for estimating species richness. The lack of contribution of the green band may be the result of the small openings within a stand. Gaps smaller than 1 ha, which are typical in the study area, are ignored in the forest stands delineation in Turkey. Therefore, the green band produces unsuitable texture values for predicting landscape metrics because it is more sensitive to these small openings than the red and infrared bands. However, this characteristic of the green band may sometimes provide an advantage depending on the topic under consideration. For example, Ozdemir et al. (2008) reported that tree size diversity at the sampling plot level was correlated with GLCM Homogeneity of the green band of ASTER image with 15 m resolution.

The image-based explanatory variables investigated in this study were able to explain considerable amount of variance (50-53%) in particular landscape metrics. However the unexplained variance is due to unknown causes. The main reason may be due to the heterogeneous stands. Most stands except plantations and naturally generated stands cannot be completely delineated as homogeneous area in terms of species or diameter class because of the complex structure of Mediterranean forests. For that reason, the landscape metrics calculated based on the stand boundaries are prone to error. A heterogeneous stand delineated as one polygon can incorporate small tree groups that might be recognized as several polygons in carrying out the multi-resolution segmentation algorithm to ASTER imagery. Furthermore, while an area composed of the heterogeneous stands seems structurally simple in the forest map, its corresponding ASTER image is highly textured. Consequently, sample windows having several heterogeneous stands yield the landscape metric values that are less correlated with the image derived features. In order to reduce this effect, more detailed forest maps are needed for assessment instead of the timber management map. These maps can be obtained by improving traditional forest maps using higher resolution satellite images (e.g. Ikonos, Quickbird). Therefore, we expect that the landscape metrics calculated based on an edited stand map might be predicted more accurately by the image parameters of ASTER.

The landscape metrics alone, which were calculated based on forest patchiness (spatial structure), may not be sufficient for quantifying forest structural diversity. In this sense, a new index that will take into account both the forest patchiness and the vertical structure (e.g. tree size diversity) should be developed. Thus, the fit of satellite-based model might be improved if the structural diversity could be defined using such a combined index. Hence, a future study integrating horizontal and vertical structure is needed to understand more clearly the potential of satellite data for modeling forest structural diversity. Any future study should also focus on satellite imageries with different spatial resolutions.

4. Conclusions

1. It can be concluded that it is possible to use the ASTER data to estimate spatial structure of forest landscapes with a reasonable accuracy at large scales. However, the ASTER spectral bands with 15 m resolution cannot replace aerial photos or very high resolution satellite imagery to predict landscape diversity in decision-making regarding forest management at local level.

2. The texture measures calculated from the red (0.63-0.69 [micro]m) and near infrared (0.76-0.86 [micro]m) bands of ASTER imagery (B group variables in this study) have a great potential for characterizing landscape spatial structure. Our results suggest that image texture may be a costeffective way of mapping the landscape heterogeneity in the Mediterranean forest landscapes.

3. The other conclusion that can be drawn from this study is that the landscape extents between approximately 81 and 144 ha may be convenient to compute the texture features in order to predict the landscape diversity in the Mediterranean forest ecosystems.

4. The multi-resolution segmentation was also a promising tool for delineating land cover polygons that were employed as the base in calculating the landscape metrics (A group variables). Furthermore, the object-based approach might address some problems with the use of categorical data. For example, forest stand maps might not show the real forest vegetation polygons since many small landscape elements are potentially merged as one stand. In contrast, landscape patches can be accurately delineated using appropriate satellite data and the multi-resolution segmentation algorithm.

doi: 10.3846/16486897.2012.688371


The authors would like to thank Prof. Dr Keith C. Clarke and Dr Stephen McNeill for their valuable comments on an early version of the manuscript and for revising the English.


Addink, E. A.; Van Coillie, F. M. B.; de Jong, S.M. 2012. Introduction to the GEOBIA 2010 special issue: From pixels to geographic objects in remote sensing image analysis, International Journal of Applied Earth Observation and Geoinformation 15(SI): 1-6.

Baatz, M.; Benz, U.; Dehgani, S.; Heynan, M.; Holtje, A.; Hofmann, P.; Lingenfelder, I.; Mimler, M.; Sohlbach, M.; Weber, M.; Willhauck, G. 2001. eCognition Obfect Oriented Image Analysis User Guide. Munich. 483 p.

Baskent, E. Z.; Jordan, G. A. 1995. Characterizing Spatial Structure of Forest Landscapes, Canadian Journal of Forest Research-Revue Comedienne De Recherche Forestiere 25(11): 1830-1849.

Baskent, E. Z.; Jordan, G. A. 1996. Designing forest management to control spatial structure of landscapes, Landscape and Urban Planning 34(1): 55-74.

Baskent, E. Z. 1997. Assessment of structural dynamics in forest landscape management, Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 27(10): 1675-1684.

Baskent, E. Z. 1999. Controlling spatial structure of forested landscapes: a case study towards landscape management, Landscape Ecology 14(1): 83-97.

Baskent, E. Z.; Terzioglu, S.; Baskaya, S. 2008. Developing and implementing multiple-use forest management planning in Turkey, Environmental Management 42(1): 37-48.

Bauza, D. 2007. Estimation and trends of landscape transformation in the second half of the 20th century, Journal of Environmental Engineering and Landscape Management 15(2): 119-124.

Blaschke, T. 2010. Object based image analysis for remote sensing, ISPRS Journal of Photogrammetry and Remote Sensing 65(1): 2-16.

Chang, A.; Eo, Y.; Kim, S.; Kim, Y.; Kim, Y. 2011. Canopy-cover thematic-map generation for Military Map products using remote sensing data in inaccessible areas, Landscape and Ecological Engineering 7(2): 263-274.

Chen, G.; Hay, G. J.; St-Onge, B. 2012. A GEOBIA framework to estimate forest parameters from lidar transects, Quickbird imagery and machine learning: A case study in Quebec, Canada, International Journal of Applied Earth Observation and Geoinformation 15(SI): 28-37.

Dye, M.; Mutanga, O.; Ismail, R. 2008. Detecting the severity of woodwasp, Sirex noctilio, infestation in a pine plantation in KwaZulu-Natal, South Africa, using texture measures calculated from high spatial resolution imagery, African Entomology 16(2): 263-275.

Egbert, S.; Park, S.; Price, K.; Lee, R.; Wu, J.; Nellis, A. 2002.

Using conservation reserve program maps derived from satellite imagery to characterize landscape structure, Computers and Electronics in Agriculture 37(1-3): 141-156.

Eiden, G.; Kayadjanian, M.; Vidal, C. 2000. Capturing landscape structures: Tools From Land Cover to Landscape Diversity. Available from Internet:

Fortin, M. J.; Melles, S. J. 2009. Avian Spatial Responses to Forest Spatial Heterogeneity at the Landscape Level: Conceptual and Statistical Challenges, in Real World Ecology--Large-Scale and Long-Term Case Studies and Methods, Eds. ShiLi Miao, Susan Carstenn and Martha Nungesser, Springer New York, 137-160.

Goetz, S.; Steinberg, D.; Dubayah, R.; Blair, B. 2007. Laser remote sensing of canopy habitat heterogeneity as a predictor of bird species richness in an eastern temperate forest, USA, Remote Sensing of Environment 108(3): 254-263.

Gu, Z. J.; Zeng, Z. Y.; Shi, X. Z.; Li, L.; Yu, D. S.; Zheng, W.; Hu, Z. F. 2009. Assessing actors influencing vegetation coverage calculation with remote sensing imagery, International Journal of Remote Sensing 30(10): 2479-2489.

Gustafson, E. J. 1998. Quantifying Landscape Spatial Pattern: What Is the State of the Art?, Ecosystems 1(2): 143-156.

Hall-Beyer, M. 2007. GLCM Texture: A Tutorial. Available from Internet:

Harralick, R. M.; Shanmugam, K.; Dinstein, I. 1973. Textural features for image classification, in IEEE Transactions on Systems, Man, and Cybernetics. SMC-3, 610-621.

Harris, L. D.; Smith, W. H. 1978. Relations of Forest practice to non-timber values and adjacent ecosystems, in Proceedings Symposium on maintaining Site Productivity Prepared Sites. Mississippi State University, Starkville, 28-53.

Hyde, P.; Dubayah, R.; Walker, W.; Blair, J. B.; Hofton, M.; Hunsaker, C. 2006. Mapping forest structure for wildlife habitat analysis using multi-sensor (LiDAR, SAR/InSAR, ETM+, Quickbird) synergy, Remote Sensing of Environment 102(1-2): 63-73.

Kadiogullari, A. I.; Baskent, E. Z. 2008. Spatial and temporal dynamics of land use pattern in Eastern Turkey: a case study in Gumushane, Environmental Monitoring and Assessment 138(1-3): 289-303.

Kim, M.; Warner, T. A.; Madden, M.; Atkinson, D. S. 2011. Multiscale GEOBIA with very high spatial resolution digital aerial imagery: scale, texture and image objects, International Journal of Remote Sensing 32(10): 2825-2850.

Kucas, A. 2010. Location Prioritization by means of Multicriteria Spatial Decision-Support Systems: A Case Study of Forest Fragmentation-Based Ranking of Forest Administrative Areas, Journal of Environmental Engineering and Landscape Management 18(4): 312-320.

Lamonaca, A.; Corona, P.; Barbati, A. 2008. Exploring forest structural complexity by multi-scale segmentation of VHR imagery, Remote Sensing of Environment 112(6): 2839-2849.

Levin, N.; Shmida, A.; Levanoni, O.; Tamari, H.; Kark, S. 2007, Predicting mountain plant richness and rarity from space using satellite-derived vegetation indices, Diversity and Distributions 13(6): 692-703.

Lillesand, T. M.; Kiefer, R. W. 1994. Remote Sensing and Image Interpretation, John Wiley & Sons Inc, 3, 766.

McGarigal, K.; Cushman, S. A.; Neel, M. C.; Ene, E. 2002. FRAGSTATS: Spatial Pattern Analysis Program for Categorical Maps. Computer software program produced by the authors at the University of Massachusetts, Amherst. Available from Internet:

Moore, D. S.; McCabe, G. P. 1993. Introduction to the Practice of Statistics. 2nd edition. 854 p. W. H. Freeman and Company, New York.

Morgan, J. L.; Gergel, S. E. 2010. Quantifying historic landscape heterogeneity from aerial photographs using object-based analysis, Landscape Ecology 25(7): 985-998.

Ota, T.; Mizoue, N.; Yoshida, S. 2011. Influence of using texture information in remote sensed data on the accuracy of forest type classification at different levels of spatial resolution, Journal of Forest Research 16(6): 432-437.

Ozdemir, I.; Norton, D. A.; Ozkan, U. Y.; Mert, A.; Senturk, O. 2008. Estimation of tree size diversity using object oriented texture analysis and ASTER imagery, Sensors 8(8): 4709-4724.

Ozdemir, I.; Karnieli, A. 2011. Predicting forest structural parameters using the image texture derived from WorldView-2 multispectral imagery in a dryland forest, Israel., International Journal of Applied Earth Observation and Geoinformation 13(5): 701-710.

Ozkan, K.; Mert, A. 2011. Ecological Land Classification and Mapping of Yazili Canyon Nature Park in the Mediterranean Region, Journal of Environmental Engineering and Landscape Management 19(4): 296-303.

Peters, D. P. C.; Goslee, S. C. 2000. Landscape Diversity, in Levin, S. (Ed.). Encyclopedia of Biodiversity. Academic Press, San Diego, CA.

Rocchini, D. 2007. Distance decay in spectral space in analysing ecosystem beta-diversity, International Journal of Remote Sensing 28(11): 2635-2644.

Rocchini, D.; Ricotta, C.; Chiarucci, A. 2007. Using satellite imagery to assess plant species richness: The role of multispectral systems, Applied Vegetation Science 10(3): 325-331.

Rocchini, D. 2009. Algorithmic Foundation of Spectral Rarefaction for Measuring Satellite Imagery Heterogeneity at Multiple Spatial Scales, Sensors 9(1): 303-310.

Rocchini, D.; Ricotta, C.; Chiarucci, A.; de Dominicis, V.; Cirillo, I.; Maccherini, S. 2009. Relating spectral and species diversity through rarefaction curves, International Journal of Remote Sensing 30(10): 2705-2711.

Schindler, S.; Poirazidis, K.; Wrbka, T. 2008. Towards a core set of landscape metrics for biodiversity assessments: A case study from Dadia National Park, Greece, Ecological Indicators 8(5): 502-514.

Smith, H. Y. 2000. Wildlife Habitat Considerations, The Bitterroot Ecosystem Management Research Project: What we have learned-symposium proceedings; 1999 May 18-20; Missoula, MT. Proceedings RMRS-P-17. Ogden, UT: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station.

St-Louis, V.; Pidgeon, A. M.; Radeloff, V. C.; Hawbaker, T. J.; Clayton, M. K. 2006. High-resolution image texture as a predictor of bird species richness, Remote Sensing of Environment 105(4): 299-312.

Turner, M. G.; O'Neill, R. V.; Gardner, R. H.; Milne, B. T. 1989. Effects of changing spatial scale on the analysis of landscape pattern, Landscape Ecology 3: 153-162.

Vega, C.; Durrieu, S. 2011. Multi-level filtering segmentation to measure individual tree parameters based on Lidar data: Application to a mountainous forest with heterogeneous stands, International Journal of Applied Earth Observation and Geoinformation 13(4): 646-656.

Ibrahim OZDEMIR. Associate Professor, Doctor at forest management department, Dr since 2003, 5 peer-reviewed papers in the journal covered by SCI, 10 peer-reviewed papers in the Turkish journals. Research interests: landscape management, forest inventory, remote sensing.

Ahmet MERT. Lecturer, Master of forest science since 2006, 4 peer-reviewed papers in the journal covered by SCI, 1 peer-reviewed paper in the journals (Turkish). Research interests: Landscape Ecology, Remote Sensing, Geographical Information System.

Ozdemir SENTURK. PhD student, Master of forest science since 2008, 2 peer-reviewed papers in the journal covered by SCI. Research interests: Landscape Ecology, Geographical Information System.

Ibrahim Ozdemir (1), Ahmet Mert, Ozdemir Senturk

Forest Management Department, Forestry Faculty, Suleyman Demirel University, Isparta, Turkey

E-mail: (1) (corresponding author)

Submitted 14 Dec. 2010; accepted 18 Dec. 2011
Table 1. The best fitted regression models

Land.     Landscape                   Equation
Metric     Extents

NUMP     1200 x 1200   = -32.073 + 1.152 SB-PR + 0.446 SB-IJI
         1200 x 1200   = -42.894 + 10.542 GENB3 - 32.084 GHOB2

ED       1200 x 1200   = 29.348 + 0.607 SB-ED
         900 x 900     = -165.546 + 23.09 GENB2 + 21.895 GENB3

IJI      900 x 900     = 37.449 + 11.211 SB-SDI

SDI      900 x 900     = -0.463 + 0.895 SB-SDI
         1200 x 1200   = - 4.747 + 0.55 GENB3 + 0.508 GENB2 -
                         0.044 GSDB2

SEI      900 x 900     = 0.12 + 0.273 SB-SDI
         1500 x 1500   = -0.114+ 0.09 GENB2

PR       1200 x 1200   = -0.784 + 0.641 SB-PR
         1200 x 1200   = -40.135 + 5.3 32 GENB3 + 2.449 GENB2

Land.    The Group of     Adj.      [S.sub.res]     p
Metric   Independent    [R.sup.2]

NUMP          A           0.53         4.04       <0.01
              B           0.51         4.13       <0.01

ED            A           0.52         14.02      <0.01
              B           0.43         19.31      <0.01

IJI           A           0.12         8.81       <0.01

SDI           A           0.44         0.31       <0.01
              B           0.52         0.26       <0.01

SEI           A           0.30         0.13       <0.01
              B           0.25         0.08       <0.01

PR            A           0.50         2.57       <0.01
              B           0.52         2.52       <0.01
COPYRIGHT 2012 Vilnius Gediminas Technical University
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2012 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Ozdemir, Ibrahim; Mert, Ahmet; Senturk, Ozdemir
Publication:Journal of Environmental Engineering and Landscape Management
Article Type:Abstract
Geographic Code:7TURK
Date:Jun 1, 2012
Previous Article:Determination of organic matter by UV absorption in the ground water/Organiniu medziagu pozeminiame vandenyje nustatymas taikant UV bangu absorbcija.
Next Article:Visit of Dr Gregory Nathaniel Taff at Vilnius Gediminas Technical University.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters