Assessing Future
hydrological response of an urban watershed using machine learning based LULC forecasting
models
Rajput Preeti, Sinha Manish Kumar and Taram Ramchandra
Disaster Advances; Vol. 17(11); 35-48;
doi: https://doi.org/10.25303/1711da035048; (2024)
Abstract
Urbanization in terms of land-use/cover (LULC) change has a long-term significant
impact on the hydrological cycle as the LULC is one of the most important influencing
parameters to produce curve number (CN). The drastic change in LULC changes the
CN. This change directly affects surface water including peak flows. This study
aims to assess the change in surface runoff due to changes in LULC. Hydrological
modeling is done for the consistent long-term behavioral study of surface runoff.
The area focused on the study is the Kharun river, a tributary of the Mahanadi River.
For the assessment of the impact of LULC change on the catchment discharge, a daily-step
conceptual model soil and water assessment tool (SWAT) was applied. Landuse maps
were prepared with the Landsat Thematic Mapper satellite images.
The future land use was forecasted with the technique of spatial statistical modeling.
The machine learning (ML) tool is used for quantifying the influences on the LULC
change dynamics and producing the LULC map for 2024 and 2030. Remote sensing (RS)
and geographic information system (GIS) analysis were coupled with hydrological
SWAT modeling to investigate the connection between the LULC change and hydrologic
regime. The SWAT Model’s calibration efficiency is verified by comparing the simulated
and observed discharge time series at the Patharidih gauge and discharge station.
The monthly and daily calibrations were quite satisfactory, with Nash-Sutcliffe
an efficiency coefficient of 0.86 and 0.67. This modeling provides reliable information
for sustainable management of available water resources of the catchment.
