Bilingual
En/Fa
Geographical Research is Published in both Persian and English Full-text.
Volume 37, Issue 3 (2022)
GeoRes 2022, 37(3): 381-389 |
Back to browse issues page
Article Type:
Subject:
History
Received: 2022/04/21 | Accepted: 2022/06/17 | Published: 2022/07/1
Received: 2022/04/21 | Accepted: 2022/06/17 | Published: 2022/07/1
How to cite this article
Yaghoobi Bayekolaee M, Vafaeenejad A, Moradi Darabkalayi H, Hashemi H. Prediction of Land Use/Cover Changes in the Gorganrood Watershed Using Metrics and Land Change Processes. GeoRes 2022; 37 (3) :381-389
URL: http://georesearch.ir/article-1-1329-en.html
URL: http://georesearch.ir/article-1-1329-en.html
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Rights and permissions
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
1- Department of Water Resources Management Engineering, Faculty of Civil, Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran
2- Department of Geotechnical and Transportation Engineering, Faculty of Civil, Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran
3- Department of Watershed Management and Engineering, College of Natural Resources, Tarbiat Modares University, Tehran, Iran
4- Department of Water Resources Engineering and Center for Advanced Middle Eastern Studies, Lund University, Lund, Sweden
2- Department of Geotechnical and Transportation Engineering, Faculty of Civil, Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran
3- Department of Watershed Management and Engineering, College of Natural Resources, Tarbiat Modares University, Tehran, Iran
4- Department of Water Resources Engineering and Center for Advanced Middle Eastern Studies, Lund University, Lund, Sweden
Full-Text (HTML) (101 Views)
Introduction
Landscape structure refers to an arrangement in which a combination of local ecosystems or land uses is repeated across a region in similar forms (Apan et al., 2002). Understanding the types of land use/land cover and human activities across different parts of a region, namely, how land is utilized serves as essential baseline information for various planning purposes (Matsushita et al., 2006). Land use/land cover is one of the key parameters of landscape structure and represents the outcome of interactions between human and natural environments. Quantifying land use/land cover heterogeneity is crucial for clarifying the relationship between spatial patterns and the occurrence of diverse natural processes, making it an important component of landscape studies (Braimoh et al., 2006).
Accordingly, examining, analyzing, and interpreting spatial patterns of land use/land cover can support proper modeling of natural processes, ultimately contributing to effective landscape management. The use of landscape metrics provides an appropriate method for quantifying landscape structure and enhances the understanding of relationships between landscape patterns and ecological processes. A wide array of landscape metrics is currently available for examining the relationships between spatial structure and ecological functions, necessitating the careful selection of the most appropriate metrics for policymaking and decision-making. The selection of suitable metrics depends on the objectives of the study, landscape characteristics, and the nature of the ecological processes involved (Buyantuyev et al., 2009). Therefore, assessing landscape change requires selecting appropriate metrics in accordance with project objectives.
Landscape metrics can be categorized into three levels (Uuemaa et al., 2001): patch-level metrics, which describe individual patches and identify their spatial characteristics, content, and texture; class-level metrics, which apply to all patches of a given type, where a class refers to all patches representing the same land use or land use/land cover category; and landscape-level metrics, which provide an integrated description of all classes and patches within the landscape (Buyantuyev et al., 2009).
In some cases, establishing direct relationships between landscape metrics and natural watershed processes is not straightforward. Behavioral changes in land use/land cover may not be adequately represented by a single or limited number of landscape metrics, as a combination of interacting factors may be involved (Fiener et al., 2011; Su et al., 2014). Limited understanding of how changes in landscape structure and configuration influence ecosystem functions remains a concern among researchers and managers (Tlapáková et al., 2013). Predicting landscape patterns based on observed change processes provides a roadmap for understanding future land use/land cover conditions and can significantly support decision-makers.
Numerous studies have investigated landscape behavior and its implications for various ecological and environmental processes. Toosi Bihamta et al. (2014) have analyzed changes in the shape and size of land-use patches in the city of Isfahan using landscape metrics and concluded that alterations in spatial characteristics influence ecological functioning and should be considered in land-use planning. Kiyani (2015), analyzing land-use structure in the Sefidrud watershed using nine landscape metrics, have found that agricultural land had experienced the highest degree of fragmentation. Joorabian Shooshtari et al. (2017), examining the role of landscape metrics in evaluating the performance of the GEOMOD model in the Nekarud watershed (Mazandaran Province), have reported that this approach provides a comprehensive understanding of uncertainty in the accuracy of model outputs. Zabihi et al. (2019) have assessed the effects of possible land use/land cover change scenarios on landscape metrics in the Talar watershed and found decreasing trends in forest and rangeland classes for metrics such as class percentage, number of patches, edge density, patch density, largest patch index, and landscape shape index, while other land uses exhibited increasing trends. Karami and Mirsanjari (2018) have reported that, over a 25-year period in the Hoveyzeh wetland region, the number of patches, patch density, largest patch index, and patch shape index decreased, whereas Shannon’s diversity index increased. Sun and Zhou (2016) have demonstrated the effectiveness of landscape metrics, including the contagion and adjacency indices in monitoring temporal and spatial changes in agricultural landscapes in China at the watershed scale. Boongaling et al. (2018) have shown that increases in patch density and largest patch index for forest and agricultural lands in the Kalumpang watershed (Philippines) led to reductions in surface runoff and sediment yield. Zabihi et al. (2020) have evaluated landscape management through monitoring change processes in the Talar watershed, focusing on Landscape Change Processes (LCP).
Despite numerous studies on landscape metrics with varying objectives, a comprehensive assessment of the effects of different possible land use/land cover change scenarios on landscape metrics has yet to be conducted. The present study aims to predict land use/land cover maps under two scenarios for the year 2040: (1) continuation of current change trends and (2) land use/land cover changes based on ecological suitability, using the Land Change Modeler (LCM) and the Weighted Linear Combination (WLC) method in the Gorganrud watershed in Golestan Province. Another objective is to examine the effects of possible land use/land cover change states on landscape metrics and landscape change processes.
Methodology
To prepare the land use/land cover maps for the Gorganrud watershed, TM sensor images from Landsat for June 1989, July 2000, July 2010, and OLI sensor imagery for June 2019 were obtained from the United States Geological Survey. Atmospheric and radiometric corrections were applied using the dark object subtraction method and the FLAASH module in ENVI software. In the dark object subtraction method, a constant value is subtracted from the digital numbers of each band. FLAASH performs atmospheric correction for visible, infrared, and near-infrared wavelengths up to 3 micrometers and provides advantages for classifying multispectral images (Feizizadeh et al., 2016).
To improve feature identification in TM and OLI images, field surveys, Google Earth, and true- and false-color composites were used to collect training samples (at least 50 per land-use class). Training samples for 1989 and 2000 were selected from locations that remained unchanged across the study period. Five land-use classes including rangeland, agriculture, forest, urban areas, and water bodies were identified. The Support Vector Machine algorithm was employed for satellite image classification. Accuracy assessment was conducted by comparing 30% of field validation points with the classified maps using the Kappa coefficient and overall accuracy. Accuracy for earlier maps (1989 and 2000) was assessed through visual interpretation and stable reference points over time (Joorabian Shooshtari et al., 2018; Yousefi et al., 2014).
The study area covers the Gorganrud watershed upstream of the Voshmgir Dam, with an area of 7,138 km², located in northern Iran between longitudes 54°42' to 56°28' E and latitudes 36°43' to 37°49' N. The highest elevation, 2,898 m, is in the Khosh-Yeylaq region in the southwest, while the lowest point, 10 m above sea level, is at the Voshmgir Dam. The average elevation of the watershed is 890 m, and the mean slope is 18% (Saffari et al., 2018). The main river, the Gorganrud, is approximately 333 km long, originating at an elevation of 2,297 m. Annual precipitation ranges from 195.2 to 946.3 mm. Mean annual temperatures range from 11°C to 18.1°C across different climatic stations.
Future land use/land cover changes were modeled under two scenarios: (1) continuation of the current trend up to 2040 and (2) land use/land cover changes in 2040 based on ecological suitability. First, the land use map for 2040 was produced using the Land Change Modeler (LCM) for the first scenario. Then, ecological suitability for potential change areas was determined, and the most suitable 10% of change-prone lands were integrated into the 2020 land use map for the second scenario (Rajaei et al., 2017). Land use/land cover maps for 1989–2000, 2000–2010, and 2010–2019 were introduced to LCM for change detection. The prediction period was selected as 2020–2040, considering the significant and likely changes occurring in the watershed. Land-use change modeling was conducted in three stages of change analysis, transition potential modeling, and future prediction using TerrSet software.
Parameters influencing land use/land cover change were selected based on previous research (Mishra et al., 2014) and data availability. These parameters were evaluated for transition potential modeling using Cramer’s V correlation coefficient. Values greater than 0.15 were considered acceptable (Reddy et al., 2017). The parameters included slope, elevation, distance to forest, residential areas, agriculture, rivers, roads, and empirical probability of change.
The multilayer perceptron (MLP) neural network was used as a robust and widely applied method for transition potential modeling (Megahed et al., 2015; Romano et al., 2018). Accuracy was evaluated through training and testing errors and overall accuracy. Future land-use prediction was then performed using the Markov chain method (Rodríguez Eraso et al., 2012), and validation was conducted through statistical and visual assessments, including various Kappa coefficients and the skill measure (Ahmadi Nadoushan et al., 2012).
In the second scenario, land use/land cover change was modeled based on ecological suitability using the Weighted Linear Combination (WLC) method within the framework of Iranian ecological models. Suitability maps were generated for the three major land-use types of forest, rangeland, and agriculture which experienced the greatest changes during the study period. Then, 10% of the most suitable areas for each class expected to undergo change by 2040 were allocated. This method identifies locations with the highest suitability that together represent 10% of the expected change. Multiple criteria were evaluated following the Multi-Criteria Evaluation (MCE) process (Rajaei et al., 2017), including:
Findings
The results of land change modeling indicate substantial transitions among the three major land-use classes—agriculture, rangeland, and forest—between 1990 and 2020. Land-use maps for 1990 to 2010, 2010 to 2020, and 2020 to 2040 were used respectively as the calibration, validation, and prediction periods.
The Cramer’s V correlation coefficients for parameters influencing land-use change during the calibration and validation periods show acceptable levels of association, confirming their suitability for transition potential modeling. Parameters with the highest correlations include distance to forest, empirical probability of change, and elevation.
The evaluation of the MLP neural network used for transition potential modeling demonstrates satisfactory performance. Accuracy, training RMS, testing RMS, and skill measure for both calibration and validation periods show that the model effectively captures the spatial patterns of land-use transitions.
Predicted land-use maps for 2040 under the trend-continuation scenario and the ecological suitability scenario were generated. These maps reflect distinct spatial patterns of future land allocation resulting from the two modeling approaches.
Landscape metrics calculated for 1990, 2000, 2010, and 2020, as well as for the two scenarios for 2040, illustrate considerable temporal variation in the structural characteristics of the watershed’s landscape. Indicators such as the number of patches, patch density, largest patch index, edge density, landscape shape index, and area-weighted mean shape index display notable shifts, reflecting increasing fragmentation and landscape complexity under the future scenarios.
The extraction of landscape change processes across the study periods shows how each land-use class has undergone transitions such as area expansion, contraction, fragmentation, coalescence, and other structural modifications. These processes collectively reveal the dynamic nature of landscape transformation in the Gorganrud watershed over the historical and projected periods.
Discussion
The analysis of land use/land cover (LULC) changes in the Gorganrud watershed indicates that over the study period (1990–2020), the most extensive alterations occurred in the degradation of forest and rangeland cover (412.6 and 279.53 km², respectively), while the greatest increase was observed in agricultural land (542.98 km²). Moreover, based on the predicted LULC map for 2040 under the two management scenarios, the areas of forest, agricultural, and rangeland uses in Scenario 1 (continuation of current trends) will reach 1,364.98, 2,396.09, and 3,481.18 km², corresponding to changes of –58.37, +35.8, and +8.28 km², respectively. In this regard, the decrease in forest area and the expansion of agricultural, rangeland, and residential areas suggest that anthropogenic factors will play a significant role in LULC change in the Gorganrud watershed. Therefore, appropriate land-use management strategies, such as forest conservation and afforestation, protection of aquatic and rangeland ecosystems, restricting agricultural expansion on steep slopes, and preventing orchard development in upstream rangelands should be adopted to ensure sustainable watershed management. Conversely, under Scenario 2 (based on ecological suitability), the areas of forest, agricultural, and rangeland uses will reach 1,427.54, 2,258.55, and 3,567.49 km², with respective changes of +4.27, –100.86, and +96.58 km².
According to the results of Cramer’s V coefficient calculated for the study periods (1990–2010 and 2010–2020), the highest value (0.41) was attributed to the parameter “distance from forest edge,” while the lowest value (0.18) corresponded to “distance from roads.” Similar findings were reported for the Nekarud watershed by Shooshtari and Gholamalifard (2015). In addition, elevation showed a Cramer’s V value of 0.33, indicating its important role in modeling transition potential, consistent with the findings of Kavian et al. (2017) in the Haraz watershed in Mazandaran Province.
Based on the results presented in the section on landscape metrics and processes, the metrics ED, LSI, NP, and PD exhibited an increasing trend from 1990 to 2010, followed by a decreasing trend from 2010 to 2020 at the landscape level in the Gorganrud watershed. The PA metric and LPI index showed a decreasing trend from 1990 to 2020. The AWMSI metric demonstrated both decreasing and increasing fluctuations over the study periods. Furthermore, changes in ED, LSI, NP, and PD indicate an increasing trend for future conditions under both scenarios, although the increase is more pronounced in Scenario 1. The PA metric is expected to continue its decreasing trend in both scenarios, while the LPI index will decrease under Scenario 1 and show a limited increase under Scenario 2.
The results related to landscape change processes revealed that degradation, creation, and dissection processes occurred during the study periods. Over the 30-year period, forest patches experienced degradation, rangelands underwent dissection, and agricultural and residential patches expanded through creation processes. The reduction in the number and area of forest patches led to degradation, which is considered an indicator of landscape deterioration driven by human activities (Bogaert et al., 2004). Similar observations have been reported for forest patches in Lorestan Province (Japolghy et al., 2017). The conversion of forest land into agricultural and residential uses, particularly in downstream areas and along patch edges has caused patch reduction and triggered degradation processes. Talebi Amiri et al. (2009) have emphasized that converting land with low agricultural potential results in rapid depletion of soil fertility within one to two years, rendering agricultural activities unfeasible.
The increase in patch number, which signals fragmentation, and the slight decrease in rangeland patch area resulted in the dissection process during 1990–2020. Accessibility is the starting point of landscape dynamics, and dissection processes increase accessibility, serving as an early warning indicator of landscape transformation in the study area. It should be noted that dissection, degradation, and fragmentation fall under natural landscape change processes, whereas anthropogenic processes such as aggregation and creation occur when ecological systems undergo continuous alteration (Bogaert et al., 2002). Thus, anthropogenic processes can be considered an advanced stage of natural landscape dynamics.
Creation processes in agricultural and residential patches in the Gorganrud watershed emerged through increases in both patch number and patch area during the study period. This reflects the expansion of these land uses at the expense of natural ecosystems such as forests and rangelands. The emergence of new patches and the expansion of residential areas signify further landscape degradation in the watershed (Talebi Amiri et al., 2009). Research in the Sefidrud watershed has shown that increases in patch number enhance landscape diversity, bringing patches closer together and exposing them to greater degradation. Similar evidence of creation processes in agricultural and residential land uses has been reported in Golestan National Park and the central Zagros region, where increases in patch number have driven landscape degradation (Zebardast et al., 2012; Japelaghi et al., 2019). Based on the observed LULC change trends and the predicted 2040 map under Scenario 1, it is expected that dissection processes will continue, and creation processes will occur for other land uses in the Gorganrud watershed. However, if land-use changes follow ecological suitability as outlined in Scenario 2, forest and rangeland uses will undergo creation processes, agriculture will experience dissection, and residential areas will undergo displacement. This is consistent with recent trends of villa construction and orchard expansion in mountainous and summer rangeland areas along the northern slopes of the Alborz Mountains, which must be controlled and restored to their appropriate ecological land use.
Conclusion
The analysis of landscape change processes indicates increasing fragmentation and disintegration of the landscape, particularly within land-use categories directly influenced by human activities in the Gorganrud watershed. Large patches have progressively been transformed into smaller ones due to anthropogenic pressures, reflecting a deteriorating landscape condition. Given the extensive rangeland cover and the occurrence of dissection processes during 2000–2010, which represent the initial stage in the sequence of landscape change processes, along with the projected creation processes for future conditions based on the predicted LULC maps, the need for protective measures and stricter control of land-use changes is more critical than ever. Furthermore, the persistent decline in forest land area poses a serious threat to the sustainability of these valuable resources and should serve as a strong warning to policymakers and practitioners involved in land-use and watershed management.
Acknowledgments: The authors would like to express their sincere gratitude to all officials of the Faculty of Civil, Water, and Environmental Engineering at Shahid Beheshti University for their support.
Ethical Permission: No ethical issues were reported by the authors.
Conflict of Interest: The present manuscript is part of the first author’s doctoral dissertation.
Authors’ Contributions: Yaghoobi Bayekolalee M (First Author); Introduction Writer/Principal Researcher/Discussion Writer (70%); Vafaeenejad A (Second Author); Methodologist/Discussion Writer (10%); Moradi Darabkalayi H (Third Author); Statistical Analyst/ Assistant Researcher (10%); Hashemi H (Fourth Author); Assistant Researcher (10%)
Funding: No funding was reported by the authors.
Landscape structure refers to an arrangement in which a combination of local ecosystems or land uses is repeated across a region in similar forms (Apan et al., 2002). Understanding the types of land use/land cover and human activities across different parts of a region, namely, how land is utilized serves as essential baseline information for various planning purposes (Matsushita et al., 2006). Land use/land cover is one of the key parameters of landscape structure and represents the outcome of interactions between human and natural environments. Quantifying land use/land cover heterogeneity is crucial for clarifying the relationship between spatial patterns and the occurrence of diverse natural processes, making it an important component of landscape studies (Braimoh et al., 2006).
Accordingly, examining, analyzing, and interpreting spatial patterns of land use/land cover can support proper modeling of natural processes, ultimately contributing to effective landscape management. The use of landscape metrics provides an appropriate method for quantifying landscape structure and enhances the understanding of relationships between landscape patterns and ecological processes. A wide array of landscape metrics is currently available for examining the relationships between spatial structure and ecological functions, necessitating the careful selection of the most appropriate metrics for policymaking and decision-making. The selection of suitable metrics depends on the objectives of the study, landscape characteristics, and the nature of the ecological processes involved (Buyantuyev et al., 2009). Therefore, assessing landscape change requires selecting appropriate metrics in accordance with project objectives.
Landscape metrics can be categorized into three levels (Uuemaa et al., 2001): patch-level metrics, which describe individual patches and identify their spatial characteristics, content, and texture; class-level metrics, which apply to all patches of a given type, where a class refers to all patches representing the same land use or land use/land cover category; and landscape-level metrics, which provide an integrated description of all classes and patches within the landscape (Buyantuyev et al., 2009).
In some cases, establishing direct relationships between landscape metrics and natural watershed processes is not straightforward. Behavioral changes in land use/land cover may not be adequately represented by a single or limited number of landscape metrics, as a combination of interacting factors may be involved (Fiener et al., 2011; Su et al., 2014). Limited understanding of how changes in landscape structure and configuration influence ecosystem functions remains a concern among researchers and managers (Tlapáková et al., 2013). Predicting landscape patterns based on observed change processes provides a roadmap for understanding future land use/land cover conditions and can significantly support decision-makers.
Numerous studies have investigated landscape behavior and its implications for various ecological and environmental processes. Toosi Bihamta et al. (2014) have analyzed changes in the shape and size of land-use patches in the city of Isfahan using landscape metrics and concluded that alterations in spatial characteristics influence ecological functioning and should be considered in land-use planning. Kiyani (2015), analyzing land-use structure in the Sefidrud watershed using nine landscape metrics, have found that agricultural land had experienced the highest degree of fragmentation. Joorabian Shooshtari et al. (2017), examining the role of landscape metrics in evaluating the performance of the GEOMOD model in the Nekarud watershed (Mazandaran Province), have reported that this approach provides a comprehensive understanding of uncertainty in the accuracy of model outputs. Zabihi et al. (2019) have assessed the effects of possible land use/land cover change scenarios on landscape metrics in the Talar watershed and found decreasing trends in forest and rangeland classes for metrics such as class percentage, number of patches, edge density, patch density, largest patch index, and landscape shape index, while other land uses exhibited increasing trends. Karami and Mirsanjari (2018) have reported that, over a 25-year period in the Hoveyzeh wetland region, the number of patches, patch density, largest patch index, and patch shape index decreased, whereas Shannon’s diversity index increased. Sun and Zhou (2016) have demonstrated the effectiveness of landscape metrics, including the contagion and adjacency indices in monitoring temporal and spatial changes in agricultural landscapes in China at the watershed scale. Boongaling et al. (2018) have shown that increases in patch density and largest patch index for forest and agricultural lands in the Kalumpang watershed (Philippines) led to reductions in surface runoff and sediment yield. Zabihi et al. (2020) have evaluated landscape management through monitoring change processes in the Talar watershed, focusing on Landscape Change Processes (LCP).
Despite numerous studies on landscape metrics with varying objectives, a comprehensive assessment of the effects of different possible land use/land cover change scenarios on landscape metrics has yet to be conducted. The present study aims to predict land use/land cover maps under two scenarios for the year 2040: (1) continuation of current change trends and (2) land use/land cover changes based on ecological suitability, using the Land Change Modeler (LCM) and the Weighted Linear Combination (WLC) method in the Gorganrud watershed in Golestan Province. Another objective is to examine the effects of possible land use/land cover change states on landscape metrics and landscape change processes.
Methodology
To prepare the land use/land cover maps for the Gorganrud watershed, TM sensor images from Landsat for June 1989, July 2000, July 2010, and OLI sensor imagery for June 2019 were obtained from the United States Geological Survey. Atmospheric and radiometric corrections were applied using the dark object subtraction method and the FLAASH module in ENVI software. In the dark object subtraction method, a constant value is subtracted from the digital numbers of each band. FLAASH performs atmospheric correction for visible, infrared, and near-infrared wavelengths up to 3 micrometers and provides advantages for classifying multispectral images (Feizizadeh et al., 2016).
To improve feature identification in TM and OLI images, field surveys, Google Earth, and true- and false-color composites were used to collect training samples (at least 50 per land-use class). Training samples for 1989 and 2000 were selected from locations that remained unchanged across the study period. Five land-use classes including rangeland, agriculture, forest, urban areas, and water bodies were identified. The Support Vector Machine algorithm was employed for satellite image classification. Accuracy assessment was conducted by comparing 30% of field validation points with the classified maps using the Kappa coefficient and overall accuracy. Accuracy for earlier maps (1989 and 2000) was assessed through visual interpretation and stable reference points over time (Joorabian Shooshtari et al., 2018; Yousefi et al., 2014).
The study area covers the Gorganrud watershed upstream of the Voshmgir Dam, with an area of 7,138 km², located in northern Iran between longitudes 54°42' to 56°28' E and latitudes 36°43' to 37°49' N. The highest elevation, 2,898 m, is in the Khosh-Yeylaq region in the southwest, while the lowest point, 10 m above sea level, is at the Voshmgir Dam. The average elevation of the watershed is 890 m, and the mean slope is 18% (Saffari et al., 2018). The main river, the Gorganrud, is approximately 333 km long, originating at an elevation of 2,297 m. Annual precipitation ranges from 195.2 to 946.3 mm. Mean annual temperatures range from 11°C to 18.1°C across different climatic stations.
Future land use/land cover changes were modeled under two scenarios: (1) continuation of the current trend up to 2040 and (2) land use/land cover changes in 2040 based on ecological suitability. First, the land use map for 2040 was produced using the Land Change Modeler (LCM) for the first scenario. Then, ecological suitability for potential change areas was determined, and the most suitable 10% of change-prone lands were integrated into the 2020 land use map for the second scenario (Rajaei et al., 2017). Land use/land cover maps for 1989–2000, 2000–2010, and 2010–2019 were introduced to LCM for change detection. The prediction period was selected as 2020–2040, considering the significant and likely changes occurring in the watershed. Land-use change modeling was conducted in three stages of change analysis, transition potential modeling, and future prediction using TerrSet software.
Parameters influencing land use/land cover change were selected based on previous research (Mishra et al., 2014) and data availability. These parameters were evaluated for transition potential modeling using Cramer’s V correlation coefficient. Values greater than 0.15 were considered acceptable (Reddy et al., 2017). The parameters included slope, elevation, distance to forest, residential areas, agriculture, rivers, roads, and empirical probability of change.
The multilayer perceptron (MLP) neural network was used as a robust and widely applied method for transition potential modeling (Megahed et al., 2015; Romano et al., 2018). Accuracy was evaluated through training and testing errors and overall accuracy. Future land-use prediction was then performed using the Markov chain method (Rodríguez Eraso et al., 2012), and validation was conducted through statistical and visual assessments, including various Kappa coefficients and the skill measure (Ahmadi Nadoushan et al., 2012).
In the second scenario, land use/land cover change was modeled based on ecological suitability using the Weighted Linear Combination (WLC) method within the framework of Iranian ecological models. Suitability maps were generated for the three major land-use types of forest, rangeland, and agriculture which experienced the greatest changes during the study period. Then, 10% of the most suitable areas for each class expected to undergo change by 2040 were allocated. This method identifies locations with the highest suitability that together represent 10% of the expected change. Multiple criteria were evaluated following the Multi-Criteria Evaluation (MCE) process (Rajaei et al., 2017), including:
- Defining objectives and identifying influencing criteria;
- Standardizing criteria (factors and constraints);
- Assigning weights;
- Combining criteria using WLC.
Findings
The results of land change modeling indicate substantial transitions among the three major land-use classes—agriculture, rangeland, and forest—between 1990 and 2020. Land-use maps for 1990 to 2010, 2010 to 2020, and 2020 to 2040 were used respectively as the calibration, validation, and prediction periods.
The Cramer’s V correlation coefficients for parameters influencing land-use change during the calibration and validation periods show acceptable levels of association, confirming their suitability for transition potential modeling. Parameters with the highest correlations include distance to forest, empirical probability of change, and elevation.
The evaluation of the MLP neural network used for transition potential modeling demonstrates satisfactory performance. Accuracy, training RMS, testing RMS, and skill measure for both calibration and validation periods show that the model effectively captures the spatial patterns of land-use transitions.
Predicted land-use maps for 2040 under the trend-continuation scenario and the ecological suitability scenario were generated. These maps reflect distinct spatial patterns of future land allocation resulting from the two modeling approaches.
Landscape metrics calculated for 1990, 2000, 2010, and 2020, as well as for the two scenarios for 2040, illustrate considerable temporal variation in the structural characteristics of the watershed’s landscape. Indicators such as the number of patches, patch density, largest patch index, edge density, landscape shape index, and area-weighted mean shape index display notable shifts, reflecting increasing fragmentation and landscape complexity under the future scenarios.
The extraction of landscape change processes across the study periods shows how each land-use class has undergone transitions such as area expansion, contraction, fragmentation, coalescence, and other structural modifications. These processes collectively reveal the dynamic nature of landscape transformation in the Gorganrud watershed over the historical and projected periods.
Discussion
The analysis of land use/land cover (LULC) changes in the Gorganrud watershed indicates that over the study period (1990–2020), the most extensive alterations occurred in the degradation of forest and rangeland cover (412.6 and 279.53 km², respectively), while the greatest increase was observed in agricultural land (542.98 km²). Moreover, based on the predicted LULC map for 2040 under the two management scenarios, the areas of forest, agricultural, and rangeland uses in Scenario 1 (continuation of current trends) will reach 1,364.98, 2,396.09, and 3,481.18 km², corresponding to changes of –58.37, +35.8, and +8.28 km², respectively. In this regard, the decrease in forest area and the expansion of agricultural, rangeland, and residential areas suggest that anthropogenic factors will play a significant role in LULC change in the Gorganrud watershed. Therefore, appropriate land-use management strategies, such as forest conservation and afforestation, protection of aquatic and rangeland ecosystems, restricting agricultural expansion on steep slopes, and preventing orchard development in upstream rangelands should be adopted to ensure sustainable watershed management. Conversely, under Scenario 2 (based on ecological suitability), the areas of forest, agricultural, and rangeland uses will reach 1,427.54, 2,258.55, and 3,567.49 km², with respective changes of +4.27, –100.86, and +96.58 km².
According to the results of Cramer’s V coefficient calculated for the study periods (1990–2010 and 2010–2020), the highest value (0.41) was attributed to the parameter “distance from forest edge,” while the lowest value (0.18) corresponded to “distance from roads.” Similar findings were reported for the Nekarud watershed by Shooshtari and Gholamalifard (2015). In addition, elevation showed a Cramer’s V value of 0.33, indicating its important role in modeling transition potential, consistent with the findings of Kavian et al. (2017) in the Haraz watershed in Mazandaran Province.
Based on the results presented in the section on landscape metrics and processes, the metrics ED, LSI, NP, and PD exhibited an increasing trend from 1990 to 2010, followed by a decreasing trend from 2010 to 2020 at the landscape level in the Gorganrud watershed. The PA metric and LPI index showed a decreasing trend from 1990 to 2020. The AWMSI metric demonstrated both decreasing and increasing fluctuations over the study periods. Furthermore, changes in ED, LSI, NP, and PD indicate an increasing trend for future conditions under both scenarios, although the increase is more pronounced in Scenario 1. The PA metric is expected to continue its decreasing trend in both scenarios, while the LPI index will decrease under Scenario 1 and show a limited increase under Scenario 2.
The results related to landscape change processes revealed that degradation, creation, and dissection processes occurred during the study periods. Over the 30-year period, forest patches experienced degradation, rangelands underwent dissection, and agricultural and residential patches expanded through creation processes. The reduction in the number and area of forest patches led to degradation, which is considered an indicator of landscape deterioration driven by human activities (Bogaert et al., 2004). Similar observations have been reported for forest patches in Lorestan Province (Japolghy et al., 2017). The conversion of forest land into agricultural and residential uses, particularly in downstream areas and along patch edges has caused patch reduction and triggered degradation processes. Talebi Amiri et al. (2009) have emphasized that converting land with low agricultural potential results in rapid depletion of soil fertility within one to two years, rendering agricultural activities unfeasible.
The increase in patch number, which signals fragmentation, and the slight decrease in rangeland patch area resulted in the dissection process during 1990–2020. Accessibility is the starting point of landscape dynamics, and dissection processes increase accessibility, serving as an early warning indicator of landscape transformation in the study area. It should be noted that dissection, degradation, and fragmentation fall under natural landscape change processes, whereas anthropogenic processes such as aggregation and creation occur when ecological systems undergo continuous alteration (Bogaert et al., 2002). Thus, anthropogenic processes can be considered an advanced stage of natural landscape dynamics.
Creation processes in agricultural and residential patches in the Gorganrud watershed emerged through increases in both patch number and patch area during the study period. This reflects the expansion of these land uses at the expense of natural ecosystems such as forests and rangelands. The emergence of new patches and the expansion of residential areas signify further landscape degradation in the watershed (Talebi Amiri et al., 2009). Research in the Sefidrud watershed has shown that increases in patch number enhance landscape diversity, bringing patches closer together and exposing them to greater degradation. Similar evidence of creation processes in agricultural and residential land uses has been reported in Golestan National Park and the central Zagros region, where increases in patch number have driven landscape degradation (Zebardast et al., 2012; Japelaghi et al., 2019). Based on the observed LULC change trends and the predicted 2040 map under Scenario 1, it is expected that dissection processes will continue, and creation processes will occur for other land uses in the Gorganrud watershed. However, if land-use changes follow ecological suitability as outlined in Scenario 2, forest and rangeland uses will undergo creation processes, agriculture will experience dissection, and residential areas will undergo displacement. This is consistent with recent trends of villa construction and orchard expansion in mountainous and summer rangeland areas along the northern slopes of the Alborz Mountains, which must be controlled and restored to their appropriate ecological land use.
Conclusion
The analysis of landscape change processes indicates increasing fragmentation and disintegration of the landscape, particularly within land-use categories directly influenced by human activities in the Gorganrud watershed. Large patches have progressively been transformed into smaller ones due to anthropogenic pressures, reflecting a deteriorating landscape condition. Given the extensive rangeland cover and the occurrence of dissection processes during 2000–2010, which represent the initial stage in the sequence of landscape change processes, along with the projected creation processes for future conditions based on the predicted LULC maps, the need for protective measures and stricter control of land-use changes is more critical than ever. Furthermore, the persistent decline in forest land area poses a serious threat to the sustainability of these valuable resources and should serve as a strong warning to policymakers and practitioners involved in land-use and watershed management.
Acknowledgments: The authors would like to express their sincere gratitude to all officials of the Faculty of Civil, Water, and Environmental Engineering at Shahid Beheshti University for their support.
Ethical Permission: No ethical issues were reported by the authors.
Conflict of Interest: The present manuscript is part of the first author’s doctoral dissertation.
Authors’ Contributions: Yaghoobi Bayekolalee M (First Author); Introduction Writer/Principal Researcher/Discussion Writer (70%); Vafaeenejad A (Second Author); Methodologist/Discussion Writer (10%); Moradi Darabkalayi H (Third Author); Statistical Analyst/ Assistant Researcher (10%); Hashemi H (Fourth Author); Assistant Researcher (10%)
Funding: No funding was reported by the authors.
References
1. Ahmadi Nadoushan M, Soffianian A, Alebrahim A (2012). Predicting urban expansion in Arak metropolitan area using two land change models. World Applied Sciences Journal. 18(8):1124-1132. [Persian] [Link]
2. Apan AA, Raine SR, Paterson MS (2002). Mapping and analysis of changes in the riparian landscape structure of the lockyer valley catchment, Queensland, Australia. Landscape and Urban Planning. 59(1):43-57. [Link] [DOI:10.1016/S0169-2046(01)00246-8]
3. Bihamta Toosi N, Safianian A, Fakheran S (2014). Analysis of land cover changes in the central part of Isfahan (Iran) using landscape metrics. Iranian Journal of Applied Ecology. 6:77-88. [Persian] [Link]
4. Bogaert J, Ceulemans R, Salvador-Van Eysenrode D (2004). Decision tree algorithm for detection of spatial processes in landscape transformation. Environmental Management. 33(1):62-73. [Link] [DOI:10.1007/s00267-003-0027-0]
5. Boongaling CGK, Faustino-Eslava DV, Lansigan FP (2018). Modeling land use change impacts on hydrology and the use of landscape metrics as tools for watershed management: The case of an ungauged catchment in the Philippines. Land use policy. 72:116-128. [Link] [DOI:10.1016/j.landusepol.2017.12.042]
6. Braimoh AK (2006). Random and systematic land-cover transitions in northern Ghana. Agriculture, Ecosystems & Environment. 113(1-4):254-263. [Link] [DOI:10.1016/j.agee.2005.10.019]
7. Buyantuyev A, Wu J, Grie C (2010). Multiscale analysis of the urbanization pattern of the Phoenix metropolitan landscape of USA: Time, space and thematic resolution. Landscape and Urban Planning. 94(3-4):206-217. [Link] [DOI:10.1016/j.landurbplan.2009.10.005]
8. Feizizadeh B, Didehban K, Gholamnia K (2016). Extraction of land surface temperature (lst) based on landsat satellite images and split window algorithm study area: Mahabad catchment. Geographical Data. 25(98):171-181. [Persian] [Link]
9. Fiener P, Auerswald K, Van Oost K (2011). Spatio-temporal patterns in land use and management affecting surface runoff response of agricultural catchments-A review. Earth-Science Reviews. 106(1-2):92-104. [Link] [DOI:10.1016/j.earscirev.2011.01.004]
10. Japelaghi M, Gholamalifard M, Shayesteh K (2019). Spatio-temporal analysis and prediction of landscape patterns and change processes in the central Zagros region, Iran. Remote Sensing Applications: Society and Environment. 15:100244. [Link] [DOI:10.1016/j.rsase.2019.100244]
11. Japolghy M, Gholamalifard M, Shayesteh K (2017). Monitoring and analysis of landscape pattern of Lorestan province and its change process in GIS environment. Journal of Natural Environment. 70(1):15-36. [Persian] [Link]
12. Joorabian Shooshtari S, Shayesteh K, Gholamalifard M, Azari M, López-Moreno J (2017). The role of landscape metrics and spatial processes in performance evaluation of GEOMOD (Case study: Neka river basin). Geography and Environmental Sustainability. 7(3):63-80. [Link]
13. Joorabian Shooshtari Sh, Shayesteh K, Gholamalifard M, Azari M, López-Moreno JI (2018). Land cover change modelling in hyrcanian forests, Northern Iran: A landscape pattern and transformation analysis perspective. Cuadernos de Investigación Geográfica. 44(2):743-761. [Link] [DOI:10.18172/cig.3279]
14. Karami P, Mirsanjari M (2018). Analysis of landscape degradation in the Hawizeh wetland by using remote sensing. Journal of Wetland Ecobiology. 10(35):39-54. [Persian] [Link]
15. Kavian A, Golshan M, Abdollahi Z (2017). Flow discharge simulation based on land use change predictions. Environmental Earth Sciences. 76(588). [Link] [DOI:10.1007/s12665-017-6906-0]
16. Kiyani V, Feghhi J (2015). Investigation of cover/land use structure of sefidrod watershed by landscape ecology metrics. Journal of Environmental Science and Technology. 17(2(65):131-141. [Persian] [Link]
17. Matsushita B, Xu M, Fukushima T (2006). Characterizing the changes in landscape structure in the lake Kasumigaura Basin, Japan using a high-quality GIS dataset. Landscape and urban planning. 78(3):241-250. [Link] [DOI:10.1016/j.landurbplan.2005.08.003]
18. McGarigal K, Marks BJ (2012). Fragstats: Spatial pattern analysis program for quantifying landscape structure. Open Journal of Ecology. 2(3). [Link]
19. Megahed Y, Cabral P, Silva J, Caetano M (2015). Land cover mapping analysis and urban growth modelling using remote sensing techniques in greater cairo region-Egypt. ISPRS International Journal of Geo-Information. 4(3):1750-1769. [Link] [DOI:10.3390/ijgi4031750]
20. Mishra VN, Rai PK, Mohan K (2014). Prediction of land use changes based on land change modeler (LCM) using remote sensing: A case study of Muzaffarpur (Bihar). Environmental Science. 64(1):111-127. [Link] [DOI:10.2298/IJGI1401111M]
21. Rajaei F, Esmaili Sari A, Salman Mahiny A, Delavar M, Gholipour M, Massah Bavani A (2017). Prediction the most suitable of agricultural zones in the tajan watershed using multi criteria evaluation (mce) approach. Town and Country Planning. 9(1):111-127. [Persian] [Link]
22. Reddy CS, Singh S, Dadhwal VK, Jha CS, Rao NR, Diwakar PG (2017). Predictive modelling of the spatial pattern of past and future forest cover changes in India. Journal of Earth System Science. 126(8):137-152. [Link] [DOI:10.1007/s12040-016-0786-7]
23. Rodríguez N, Armenteras D, Retana J (2012). Land use and land cover change in the Colombian andes: Dynamics and future scenarios. Journal of Land Use Science. 8:1-21. [Link] [DOI:10.1080/1747423X.2011.650228]
24. Romano G, Abdelwahab OM, Gentile F (2018). Modeling land use changes and their impact on sediment load in a mediterranean watershed. Catena. 163:342-353. [Link] [DOI:10.1016/j.catena.2017.12.039]
25. Saffari A, Noori A, Karami J (2018). Investigation about the influence of land-cover and land use changes on soil erodibility potential, case study: Gharesou, Gorganrood. Journal of Spatial Analusis Environmental Hazards. 5(1):83-96. [Persian] [Link] [DOI:10.29252/jsaeh.5.1.83]
26. Shooshtari SJ, Gholamalifard M (2015). Scenario-based land cover change modeling and its implications for landscape pattern analysis in the Neka watershed, Iran. Remote Sensing Applications: Society and Environment. 1:1-19. [Link] [DOI:10.1016/j.rsase.2015.05.001]
27. Su S, Ma X, Xiao R (2014). Agricultural landscape pattern changes in response to urbanization at ecoregional scale. Ecological Indicators. 40:10-18. [Link] [DOI:10.1016/j.ecolind.2013.12.013]
28. Sun B, Zhou Q (2016). Expressing the spatio-temporal pattern of farmland change in arid lands using landscape metrics. Journal of Arid Environments. 124:118-127. [Link] [DOI:10.1016/j.jaridenv.2015.08.007]
29. Talebi Amiri S, Azari Dehkordi F, Sadeghi S, Soufbaf S (2009). Study on landscape degradation in Neka watershed using landscape metrics. Environmental Sciences. 6(3):133-144. [Persian] [Link]
30. Tlapáková L, Stejskalová D, Karásek P, Podhrázská J (2013). Landscape metrics as a tool for evaluation landscape structure-case study Hustopeče. European Countryside. 5(1):52-70. [Link] [DOI:10.2478/euco-2013-0004]
31. Uuemaa E, Roosaare J, Oja T, Mander Ü (2011). Analysing the spatial structure of the Estonian landscapes: Which landscape metrics are the most suitable for comparing different landscapes. Estonian Journal of Ecology. 60(1):70-80. [Link] [DOI:10.3176/eco.2011.1.06]
32. Yousefi S, Tazeh M, Mirzaee S, Moradi H, Tavangar S (2014). Comparison of different classification algorithms in satellite imagery to produce land use maps (Case study: Noor city). Journal of RS and GIS for Natural Resources (Journal of Applied RS and GIS Techniques in Natural Resource Science). 5(3):67-76. [Persian] [Link]
33. Zabihi M, Moradi H, Gholamalifard M, Khaledi Darvishan A (2019). Effects of land use/land cover change scenarios on landscape metrics on the Talar watershed. Watershed Management Researches (Pajouhesh-Va-Sazandegi). 32(1(122)):84-99. [Persian] [Link]
34. Zabihi M, Moradi H, Gholamalifard M, Khaledi Darvishan A, Fürst C (2020). Landscape anagement through change processes monitoring in Iran. Sustainability. 12(5):1753. [Link] [DOI:10.3390/su12051753]
35. Zebardast L, Yavare A, Salehi E, Makhdoum M (2012). Using landscape ecological metrics to investigate impacts of road on structural changes in Golestan national park during 1987 to 2010. Environmental Researches. 2(4):11-20. [Link]