A downscaling approach for SMAP soil moisture estimates using in situ measurements and a vegetation index.

Soil moisture monitoring at a global or regional scale is nowadays achieved through remotelysensed observations which are considered as a reliable source of information for soilmoisture. Due to the coarse spatial resolution of the available remotely sensed products, thereis need for a downscaling approach which is usually based on other finer scale remotelysensed information along with in situ soil moisture measurements. The objective ofthe downscaling approach is to produce fine resolution soil moisture estimates byexamining the possible relationships between optical / thermal remotely senseddata and commonly monitored in situ meteorological information. In this study adownscaling approach for remotely sensed land surface soil moisture observations fromNASA’s Soil Moisture Active Passive (SMAP) mission was developed. In our workwe used daily global observations acquired by the L-band (1.4 GHz) operatingmicrowave radiometer at a spatial resolution of ∼ 40 km, while the daily in situsoil moisture observations along with other meteorological parameters, like airtemperature and precipitation, were retrieved from a six-stations network located in NorthGreece. The aim of our study was to integrate SMAP passive soil moisture observationsand remotely sensed environmental information in the form of the NormalizedDifference Vegetation Index (NDVI) along with daily in situ measurements in order todevelop a downscaling approach. In consonance with the preliminary conducteddiagnostic tests, SMAP passive soil moisture retrievals, Moderated Resolution ImagingSpectroradiometer (MODIS) NDVI data, daily precipitation and daily mean air temperaturewere found to be possible predictors for soil moisture estimates at a daily time stepwith a resolution of 1 km. Thus, a multilinear regression model was established byintegrating SMAP passive Level 3 soil moisture data, Moderated Resolution ImagingSpectroradiometer (MODIS) 16-day NDVI values, daily precipitation and daily mean airtemperature values. During further examination of the developed regression equation, theperformance of our methodology was enhanced as land use information was added to theprocess. Pursuant to the initial results of the developed approach using in situ measurements for atime period of 20 months, it is argued that SMAP passive soil moisture downscaling can beeffective with the utilization of supplementary environmental information. The accuracy ofour results was ascertained by examining some fundamental statistical parameters, such asthe Residual Standard Error (RSE) and the Multiple R squared (R2). Our resultsindicated an R2value of 0.56 and a low RSE value of 0.04 m3⋅m−3 demonstratingthus the potential of our methodology to achieve soil moisture estimates at a fineresolution. The above described downscaling procedure was conducted with freely available remotesensing data, whilst data processing was executed within the R statistical computingenvironment. [ABSTRACT FROM AUTHOR]