Analysis and Prediction of Land Cover/Use Changes in Mosul City Using an integrated Approach of Object-Oriented Processing and CA-Markov Model

Introduction
Land use/land cover change is a complex process influenced by natural, economic, social, and political factors. In post-conflict regions, such changes can accelerate abnormally and generate unsustainable spatial patterns. Mosul, in northern Iraq, has experienced profound land use transformations due to rapid population growth, economic fluctuations, and most importantly, the consequences of socio-political crises and armed conflict. These changes threaten natural resources, food security, and residents' quality of life. Despite significant advances in remote sensing and GIS for land use assessment, the existing literature has rarely addressed land use changes in post-conflict cities through an integrated approach combining technical modeling precision with explicit environmental harmfulness analysis. This study fills that gap by analyzing and predicting land use/cover changes in Mosul using a hybrid approach combining object-based image analysis with a support vector machine (OBIA-SVM) and a CA-Markov model. The main research question asks whether land use changes in Mosul between 2000 and 2024 followed an unsustainable pattern and whether this trend will continue until 2050. The testable hypothesis is that land use changes in Mosul have shifted towards greater environmental harmfulness. The main innovation lies in integrating OBIA-SVM and CA-Markov with harmfulness analysis, focusing on the environmental and social consequences of changes in a post-conflict city.
 
Methodology
The methodology was implemented in four main steps. First, Landsat 5, 8, and 9 images were obtained for 2000, 2008, 2016, and 2024 at eight-year intervals. All images were Level-2 Surface Reflectance products having undergone standard atmospheric correction. Preprocessing included band stacking, study area subsetting, and contrast enhancement in ENVI 5.3. Second, object-based classification using a support vector machine (OBIA-SVM) was performed in eCognition Developer 9.01. Multi-resolution segmentation used optimized parameters: for 2000 and 2008 images, scale 65, shape 0.5, and compactness 0.5; for 2016 and 2024 images, scale 180, shape 0.7, and compactness 0.3. Spectral, textural (GLCM), geometric features, and indices including NDVI, NDBI, NDWI, SAVI, BLFEI, and BSI were extracted. Classification used an SVM with a Radial Basis Function kernel. Third, the CA-Markov model in TerrSet simulated future changes. Land use maps for 2000 and 2024 served as primary inputs. A transition probability matrix between six classes (active cropland, barren land, built-up, fallow agricultural land, green space, and water bodies) was calculated from observed changes over 24 years. For validation, the 2000-2008 matrix predicted the 2016 map, which was compared to the actual 2016 map. Using the 2000-2024 matrix and a cellular automaton with a 5×5 filter and 26 iterations, the 2050 land use map was predicted. Fourth, a harmfulness analysis calculated environmentally destructive transitions in GIS. Classification accuracy was assessed using overall accuracy and Kappa coefficient.
 
Findings
The OBIA-SVM classification achieved high accuracy, with an overall accuracy of 94% and a Kappa coefficient of 0.912, attributable to the integration of spectral, textural, and shape information. Mosul experienced severe and largely unsustainable land use changes between 2000 and 2024. Built-up areas increased from 132.38 km² in 2000 to 234.14 km² in 2024, an increase of approximately 77%. This unbalanced growth exceeds conventional urbanization trends and has been directly influenced by intensive, short-term post-crisis reconstruction programs. Active cropland decreased from 33.65 km² to 25.70 km², a reduction of about 24%. Barren land decreased by approximately 47%, from 165.92 km² to 87.81 km², indicating direct and widespread conversion of these lands to urban development during reconstruction and reflecting weak regulatory mechanisms. Water bodies decreased from 8.95 km² to 5.32 km², and green space decreased from 8.77 km² to 6.15 km². Validation of the CA-Markov model for 2016 showed an overall accuracy of 86.51% and a Kappa coefficient of 0.81, indicating satisfactory capability to simulate spatial changes. The prediction for 2050 suggests that if current trends continue, built-up areas will reach 235.03 km², while active cropland will decline to 25.71 km², water bodies to 8.58 km², and green space to 23.42 km². The harmfulness analysis reveals that approximately 24.29% of the city's area (107.16 km²) falls into the "very high harmfulness" category, primarily associated with the conversion of barren land (16.038%) and fallow agricultural land (4.862%) to built-up areas. About 62.97% of the city exhibits very low harmfulness, and 6.8% shows medium harmfulness, which could become new degradation hotspots.
 
Discussion and Conclusion
The results clearly demonstrate that land use changes in Mosul between 2000 and 2024 followed a deeply unsustainable and harmful pattern, and the predictive model indicates the continuation of this trend until 2050. The research hypothesis is confirmed. The rapid expansion of built-up areas at the expense of agricultural land, barren land, water bodies, and green space reflects the dominance of the "urgency of physical reconstruction" in the post-conflict period over any logic of sustainable planning. While this finding aligns with previous studies in arid regions, the intensity and spatial concentration of degradation in Mosul are considerably more pronounced due to its specific post-conflict conditions. The direct conversion of barren and agricultural land to urban areas indicates a severe failure of regulatory mechanisms during reconstruction. The 24% reduction in active cropland seriously threatens regional food security and increases dependence on food imports. Furthermore, the reduction in green space and water bodies contributes to urban heat island effects, diminished air quality, and increased environmental vulnerability, creating a vicious cycle of degradation. Although the CA-Markov model demonstrates acceptable capability in simulating long-term trends, its most important limitation is its inherently trend-based nature; it cannot predict sudden socio-political shocks as discrete variables. The 2050 prediction primarily assumes the continuation of driving forces observed over the past two decades. To ensure sustainable development in Mosul, adopting integrated, evidence-based policies for water resource management, protecting remaining agricultural lands, restoring urban green spaces, and replacing unregulated reconstruction with participatory land governance is an unavoidable necessity. If these challenges are ignored, Mosul will face intensified environmental crises, resource scarcity, and a severe decline in quality of life. Future studies could combine such trend-based models with qualitative scenario-building and agent-based models to achieve a deeper understanding of the root causes of these changes. Without ecological balance, post-conflict reconstruction will only accelerate Mosul's environmental collapse.