Multi-Decade Land Cover/Land Use Dynamics and Future Predictions for Zambia: 2000 - 2030

Abstract

Abstract Accurate and up-to-date information on land use/land cover change (LULCC) is important in land use planning and natural resource management; however, in sub-Saharan Africa, detailed information on LULCC is still lacking. Therefore, this study assessed the dynamics of LULC change (2000–2020) and future projections (2020–2030) for Zambia. The 2000 and 2010 LULC maps were used to simulate the 2020 LULC scenario using Artificial Neural Network (Multi-layer Perception) algorithms in Modules for Land Use Change Evaluation (MOLUSCE) plugin in QGIS 2.18.14. The 2010 and 2020 maps were used to predict the 2030 LULC classes. The reference 2020 and predicted 2020 LULC maps were used to validate the model. The validation between the predicted and observed 2020 LULC map, Kappa (loc) was 0.9869. The ANN-MLP simulated the 2020 LULC patterns successfully as indicated by the high accuracy level of more than 95%. LULC classes were predicted for 2030 using the 2010–2020 calibration period. The expected LULC types for 2030 revealed that built-up area will increase by 447.20 km2 (71.44%), while 327.80 km2 (0.73%) of cropland will be lost relative to 2020 LULC map. Dense forest (0.19%), grassland (0.85%) and bare land (1.37%) will reduce from 2020–2030. However, seasonally flooded, sparse forest, shrub land, wetland and water body will increase marginally. The largest LULC change is from forest into other LULC types. The insights from this study show that ANN-MLP can be used to predict LULCC, and that the generated information can be employed in land use planning at a national scale.