Land subsidence modelling using tree-based machine learning algorithms.

摘要: Land subsidence (LS) is among the most critical environmental problems, affecting both agricultural sustainability and urban infrastructure. Existing methods often use either simple regression models or complex hydraulic to explain predict LS. There are few studies that identify risk factors of LS using machine learning models. This study compares four tree-based for land hazard modelling at a area in Hamadan plain (Iran). The also analyzes importance six including topography (elevation, slope), geomorphology (distance from stream, drainage density), hydrology (groundwater drawdown) lithology on Thematic layers each variable related phenomenon prepared utilized as inputs models, Rule-Based Decision Tree (RBDT), Boosted Regression Trees (BRT), Classification And (CART), Random Forest (RF) algorithms produce consolidated map. accuracy generated maps then evaluated under receiver operating characteristic curve (AUC) True Skill Statistics (TSS). RF approach had lowest predictive error mapping (i.e., AUC 96.7% training, 93.8% validation, TSS 0.912 0.904 validation) followed by BRT. Groundwater drawdown was seen be influential factor contributed present area, distance stream network.