Spatial and Temporal Variability Mapping of Future Flood Hazard Affected by Climate and Land-use Changes in Kendari City, Indonesia

Abstract

Introduction The predictions of current and future flood risk in the Kendari City Southeast Sulawesi Province, Indonesia. Methods In estimating this phenomenon, 51 flood and non-flood locations were identified and mapped. A total of 20 flood risk factors were selected to model flood risk using several machine learning techniques: classification and regression tree (CART), support vector machines (SVM), multivariate discriminant analysis (MDA), and ensemble. In exploring the impact of climate change and land use changes in the future (2050) on future flood risk, the General Circulation Model (GCM) with representative concentration pathways (RCPs) of the 2.6 and 8.5 scenarios by 2050 was adopted to understand the impact on 8 variables rainfall. In addition, the CA-Markov model was also applied to future land use in 2050. To validate it, Receiver Operating Characteristic-Area Under Curve (ROC-AUC) statistical analysis and other statistical analyses were carried out. Result The ensemble model shows the performance of the AUC value with the highest prediction (AUC = 0.99) and is followed by SVM (0.99), MDA (0.97), and then CART (0.96). It is estimated that areas with moderate to very high risk of flooding will increase as a result of changes in climate and land use by 2050. As a result of these changes, areas classified as having moderate to very high-risk increase from the four models. The proportion of risk zone areas from the current distribution to 2050 in the RCP 2.6 scenario changes in the ensemble model. Very low = + 36.76%, Low=-17.14%, Moderate=-14.53%, High=-2.56%, and Very high=-2.53%. However, this change becomes more significant in the RCP 8.5 scenario from the current percentages: Very low=-0.001%, Low=-12.78%, Moderate = + 2.14%, High = + 6.12%, and Very high = + 4.52%. The results of this research can help stakeholders in disaster mitigation efforts.