Surface and Near-surface Moisture Content Assessment using Multi-Temporal Satellite Images over Perak Tengah and Manjung Regions, Malaysia

Soil moisture is considered as the most significant boundary condition controlling
precipitation, especially in the semi-arid zones. On the regional scale, the importance of soil
moisture appears in agricultural assessment (crops yield management, irrigation management,
etc.), flood and draught control. Based on these principles, the study was carried out to estimate
surface moisture content (θ) over Perk Tengah and Manjung districts in Malaysia using optical
images from multi-temporal satellites which are NOAA/AVHRR and MODIS. In order to
generate the moisture maps, “Universal Triangle” algorithm was used for NOAA/AVHRR
based on land Surface Temperature (Ts) and the Normalized Difference Vegetation Index
(NDVI) extracted from images beside field measurements of θ. θ also estimated from MODIS
through the extraction of soil wetness index (SWI) which is a sensitive parameter that controls
the mechanism of land surface and the processes at the atmosphere. Throughout the study area,
θ was measured using soil moisture probe for some parts of the study area and the oven method
for the others; both Ts and θ were measured at time of satellites overpass in two different near
surface depths 5 cm and 10 cm to examine the depth influence on θ and Ts magnitudes. The
study area was divided into three basic classes according to the nature of surface cover which
were: urban area, agricultural area and multi-types surface cover area. Moisture content maps
were generated from both satellites for each surface cover type then; generalized moisture maps
were produced through correlating the three different surface cover types for each satellite
using weightage method. Finally, two sets of validation were applied to the resultant moisture
maps. Firstly, experimental validation was performed between the satellites estimated θ and θ
values measured in-situ. Good relationships were found with R
2
reached 0.79 and 0.76 for
NOAA and MODIS sensors respectively. Secondly, a success rate curve based on spatiostatistical
technique
was
used
for
validating
the
generalized
maps
in
order
to
study
the
high-low

distribution

of θ within NOAA and MODIS generalized maps. The resultant validation
reflected a high compatibility represented by area under curve of (0.81) 80%.