Use of the classification by a decision tree in the analysis of the effect of urban dynamics on the consumption of agricultural land in the municipality of Batna
Abdelhalim Bendib,Karima Berghout
DOI: https://doi.org/10.1007/s12524-023-01702-3
IF: 1.894
2023-05-01
Journal of the Indian Society of Remote Sensing
Abstract:In the municipality of Batna, urban dynamics consume peri-urban areas, putting the most fertile agricultural land at risk. Preserving these spaces becomes a priority. However, the lack of up-to-date data may limit the scope of this mission, allowing for uncontrolled urbanization. The main objective of this study is to orient our demands toward new techniques of territorial monitoring. Remote sensing (RS) and Geographic Information Systems (GIS) were used to classify and detect changes in land cover (LC). We applied a classification by decision tree (DT) and Cellular Automata based on Artificial Neural Network (CA-ANN) model to predict future changes. Firstly, using the land cover map in 1984 and 2001 we predicted the land cover in 2019. The prediction accuracy between real and simulated land cover in 2019 was assessed using the receiver operating characteristic (ROC) Curve and Kappa statistic metrics. With 92.10% of AUC (area under the curve) and 82.55% of correctness, the CA-ANN model could better simulate the spatial pattern of 2019 land cover. Then, after calibrating and validating, the spatial distribution of land cover in 2037 was generated. The results reveal that there is a substantial rivalry between built-up area (BUA) and agricultural land (AL) between 1984 and 2019. Bare land (BL), built-up area (BUA), and steppe/dry land (St/DL) have increased by 401.39%, 250.22%, and 97.18% while agricultural land (AL), irrigated land (IL), and dense forests (DF) have decreased by 56.75%, 51.30%, and 42.11%, respectively. The simulation findings show that built-up areas (BUA), agricultural land (AL), and agricultural/steppe land (Ag/SL) may rise by 28.78%, 15.17%, and 37.60%, respectively. Irrigated land (IL), bare land (BL), steppe/dry land (St/DL), and dense forests (DF) may decrease at a high rate by 84.56%, 85.27%, 67.91%, and 51.66%, respectively, by 2037. These findings enabled us to draw up precise statistical reports that would help land use planners and policymakers in framing more adequate and coherent policies.
environmental sciences,remote sensing