The effect of urban cover fraction on the retrieval of space-borne surface soil moisture at L-band

Abstract

The world's first satellite dedicated to soil moisture measurement was launched by the European Space Agency on 2nd November 2009. One objective of this Soil Moisture and Ocean Salinity (SMOS) mission is to obtain global near-surface (top ~5 cm) soil moisture every 2 to 3 days with a target accuracy of 0.04 m 3 /m 3 . To achieve this goal, the Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) mounted on SMOS is used to observe the microwave emission from the Earth's surface at L-band (~1.4 GHz) with a spatial resolution of ~45 km. The soil moisture and vegetation water content are then retrieved simultaneously from fully-polarized brightness temperature observations at multiple incidence angles. However, given the coarse scale of the SMOS pixels, existence of non-soil targets (such as urban area, standing water, and surface rock) contributes to the overall microwave emission, potentially reducing the accuracy of the retrieved soil moisture. The impact of such confounding factors has not been thoroughly assessed in the context of soil moisture retrieval algorithms, since their microwave behaviours have not been well understood and/or modelled. Ignoring their contribution or simply assuming they behave like vegetated soil targets may result in an increased retrieval error. Therefore, microwave contribution of urban areas has been considered as a brightness temperature uncertainty for current generation soil moisture retrieval models that do not account for their presence. Consequently, this study has investigated the relationship between urban induced brightness temperature uncertainties and urban fraction under warm dry soil conditions, and then used this relationship to identify SMOS pixels expected to have non-negligible impacts from urban areas. An airborne dataset: the NAFE'06 (National Airborne Field Experiment 2006), and two urban extension datasets: the LUNSW (Land use: New South Wales) and the MODIS 500-m (MODIS Urban Land Cover 500-m) were used to make this assessment. The NAFE'06 was a large scale field campaign undertaken over the Murrumbidgee catchment, in southeast of Australia, in spring of 2006. A mean surface (top 5 cm) soil moisture of 0.07 m 3 /m 3 and a mean surface (2.5 cm) soil temperature of 21 °C was observed during the campaign. The brightness temperature observations over the city of Wagga Wagga and surrounding natural land surfaces within the Kyeamba study area from NAFE'06 were used to establish the urban induced brightness temperature uncertainties and urban fraction. The LUNSW dataset was used to distinguish urban areas and calculate urban fraction. As expected, under dry soil conditions, urban induced brightness temperature uncertainties increased with urban fraction. To assess whether urban induced brightness temperature can be ignored, the target SMOS brightness temperature error of 4K was used as a benchmark by which to set an urban fraction threshold for constraining the brightness temperature uncertainty. A urban fraction threshold of 9 - 16% was obtained, and subsequently applied to the MODIS 500-m urban fraction map in order to identify SMOS pixels across Australia that are expected to have non-negligible urban induced brightness temperature uncertainties. The critical assumption of this method is that the microwave behaviour of the studied city is similar to that of all cities in Australia. This study suggests that approximately 0.2% (116 out of 43,323) SMOS pixels over Australia are likely to have non-negligible brightness temperature uncertainties due to urban areas under warm dry conditions.