Loading…

Combining SMOS with visible and near/shortwave/thermal infrared satellite data for high resolution soil moisture estimates

•NDVI and NDWI were tested as inputs in a downscaling algorithm for SMOS L2 product.•SMOS L2 soil moisture maps were disaggregated down to 500m of spatial resolution.•The target accuracy of SMOS (∼0.04m3m−3) was achieved in the downscaled maps.•Soil moisture maps of the Iberian Peninsula were produc...

Full description

Saved in:
Bibliographic Details
Published in:Journal of hydrology (Amsterdam) 2014-08, Vol.516, p.273-283
Main Authors: Sánchez-Ruiz, Sergio, Piles, María, Sánchez, Nilda, Martínez-Fernández, José, Vall-llossera, Mercè, Camps, Adriano
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•NDVI and NDWI were tested as inputs in a downscaling algorithm for SMOS L2 product.•SMOS L2 soil moisture maps were disaggregated down to 500m of spatial resolution.•The target accuracy of SMOS (∼0.04m3m−3) was achieved in the downscaled maps.•Soil moisture maps of the Iberian Peninsula were produced.•Only readily available remotely sensed data were required. Sensors in the range of visible and near–shortwave–thermal infrared regions can be used in combination with passive microwave observations to provide soil moisture maps at much higher spatial resolution than the original resolution of current radiometers. To do so, a new downscaling algorithm ultimately based on the land surface temperature (LST) – Normalized Difference Vegetation Index (NDVI) – Brightness Temperature (TB) relationship is used, in which shortwave infrared indices are used as vegetation descriptors, instead of the more common near infrared ones. The theoretical basis of those indices, calculated as the normalized ratio of the 1240, 1640 and 2130nm shortwave infrared (SWIR) bands and the 858nm near infrared (NIR) band indicate that they are able to provide estimates of the vegetation water content. These so-called water indices extracted from MODIS products, have been used together with MODIS LST, and SMOS TB to improve the spatial resolution of ∼40km SMOS soil moisture estimates. The aim was to retrieve soil moisture maps with the same accuracy as SMOS, but at the same resolution of the MODIS dataset, i.e., 500m, which were then compared against in situ measurements from the REMEDHUS network in Spain. Results using two years of SMOS and MODIS data showed a similar performance for the four indices, with slightly better results when using the index derived from the first SWIR band. For the areal-average, a coefficient of correlation (R) of ∼0.61 and ∼0.72 for the morning and afternoon orbits, respectively, and a centered root mean square difference (cRMSD) of ∼0.04m3m−3 for both orbits was obtained. A twofold improvement of the current versions of this downscaling approach has been achieved by using more frequent and higher spatial resolution water indexes as vegetation descriptors: (1) the spatial resolution of the resulting soil moisture maps can be enhanced from ∼40km up to 500m, and (2) more accurate soil moisture maps (in terms of R and cRMSD) can be obtained, especially in periods of high vegetation activity. The results of this study support the use of high resolution LST and SWI
ISSN:0022-1694
1879-2707
DOI:10.1016/j.jhydrol.2013.12.047