Abstract:
Population growth raises demand and competition for water resources and food stocks while
it changes the landuse types by anthropogenic activities to adopt applicable measures for
supplying water for domestic, agricultural, and industrial purposes. These changes alter the
hydrological response of the river basins and can impose the communities to severe
environmental risks like floods and landslides. Therefore, understanding of landuse change is
crucial to study river basins’ behavior and take mitigatory measures. The study presented here
quantifies and analyzes the historical deforestation and landuse/landcover (LULC) change
impacts on flood peak discharge of the Maduru Oya river basin, Sri Lanka using Hydrologic
Engineering Centre-Hydrologic Modeling System (HEC-HMS) and remote sensing
techniques. The Landsat Multispectral Scanner (MSS), Thematic Mapper (TM), and
Operational Land Imager-thermal Infrared Sensor (OLI-TIRS) images are acquired in 1976,
1994, 2009, 2021 and classified using maximum likelihood algorithm of supervised
classification.
The analysis of LULC change revealed that LU change was faster and in high magnitude from
1976 to 1994 compared to the remaining period to 2021. The LULC change quantification by
analyzing each scenario revealed a 24.9% deforestation while a 2.2%, 9.8%, 8.4%, and 4.5%
increase in homestead/garden, paddy, scrubland, and water body between 1976 to 1994,
respectively. The deforestation further continued to a rate of 4.1% and a 2.0% decrease in
water bodies was also found in 2009 while homestead/garden, paddy, and scrubland continued
to increase by 3.5%, 1.4%, and 1.5% compared to 1994 landuse scenario, respectively. In
contrast, the 2021 landuse scenario indicated a 7.6% decrease in scrubland while 3.6%, 0.5%,
1.5%, and 1.8% increase in forests, homestead/garden, paddy, and water bodies. The classified
images were subjected to accuracy assessment. The overall accuracy of 82%, 84%, 88%, and
91% are found for 1976, 1994, 2009, and 2021 LU scenarios while having kappa coefficients
of 0.78, 0.80, 0.85, and 0.89 for respective years. The Normalized Difference Vegetation Index
(NDVI) assessment of scenarios corresponds to the landuse classified images.
An event-based HEC-HMS model is used to simulate the flood events in the Welikanda
catchment of the Maduru Oya river basin. The model is calibrated and validated using the
1976 landuse and then the subsequent landuses are applied to study LU change impact on flood
peak discharge. For model performance evaluation, the Nash-Sutcliffe, RMSE Observations
Standard Deviation Ratio (RSR) Percent Bias (PBIAS), and the Coefficient of determination
(R
2
) were exploited. The average NSE, RSR, PBIAS, and R
2
values of 0.92, 0.25, 17.60, and
0.94 achieved in calibration and 0.73, 0.50, -3.03, and 0.78 are found in the validation which
all can be rated very good performance except for PBIAS as satisfactory in calibration and
NSE as good in the validation. The land cover change resulted in an increase (22.3%) in flood
peak from 842 m
3
/s in 1976 to 1,030 m
3
/s in 2021. As a result of the landcover changes, the
volume is also increased (42.3%) from 178.16 MCM in 1976 to 253.52 MCM in 2021. This
study provides useful information for land and water managers, forests conservation units, and
hydrologist to understand the LULC change impacts on floods and paves the way for broad
LU and hydrological studies in Sri Lanka which are rarely conducted. The same approach can
be applied in different parts of Sri Lanka which are exposed to severe LU changes.