•  
  •  
 

Abstract

Land use in the floodplain of Demak Regency has changed rapidly under pressure from urbanization and economic activity, and this change is hypothesized to contribute to increased flood risk, a concern reinforced by the extreme flood disaster of February-March 2024, although establishing a direct causal link between the observed land-use trajectory and the magnitude of that specific flood event would require dedicated hydrological modeling beyond the scope of this study. This study proposes examining the spatio-temporal dynamics of land use from 2002 to 2024 and forecast the conditions in 2035 utilizing a Land Change Model (LCM) framework combining Markov chains with a logistic regression sub-model. Land-use classification for the periods 2002, 2013, and 2024 was obtained from Landsat 7 and 8 imageries using the Classification and Regression Trees (CART) algorithm with a classification scheme referring to the Demak Regency Spatial Plan (RTRW). Prior to modeling, candidate driving variables (elevation, slope, distance to road, distance to river and rainfall) were screened using Cramer’s V association test, distance to river was excluded (V= 0.07 and 0.01 for the 2002-2013 and 2013-2024 periods, respectively) while elevation, slope, rainfall and distance to road were retained. Hindcasting validation for 2024 achieved Kstandard= 0.887, Kno= 0.935, Klocation=0.914, KlocationStrata=0.914 and Figure of Merit (FoM) of 44.87%, values consistent with, though not exceptional relative to comparable LCM studies. The historical trend (2002-2024) demonstrated a significant transformation of coastal regions, particularly in the expansion of mangrove ecosystems. Meanwhile, the projected results for 2035 indicate a continued expansion of built-up areas (+536.31 ha) and an ongoing vulnerability of the economic sector to coastal dynamics (fisheries zone decreasing by -1,037.64 ha). In contrast, mangrove ecosystems and food agricultural lands are projected to remain relatively stable from 2024 to 2035. These results should be read as model-based projections conditional on the 2013-2024 transition regime, not as a deterministic forecast, given the non-stationary pressures (tidal flooding, land subsidence, river normalization, and mangrove rehabilitation programs) operating in this coastal lowland. Cramer’s V is used here as a screening step to reduce the driver set to variables with meaningful categorical association to land-use change, it does not by itself establish the absence of multicollinearity among predictors, and this distinction is treated explicitly in the discussion. The resulting driver set and prediction map are intended as one input, among others, for flood hazard modeling and spatial planning evaluation in tropical flood-prone watersheds.

First Page

158

Last Page

178

Download

Share

COinS