Towards a daily soil moisture product based on incomplete time series observations of two satellites

Abstract

There is a strong need for a reliable daily continental soil moisture product that captures water stored in the top soil layer of the land surface and available for evaporation. Soil moisture is an important component of the water balance in hydrological systems. It has been shown that updating soil moisture at daily intervals in a numerical weather model leads to an increase in precipitation forecast skill. Better daily soil moisture has potential for improved hydrological models and improved prediction of water availability, runoff, flood discharge, etc. Furthermore, uncertainty information about soil moisture data will enable hydrologists, ecologists, agriculturalists, and climatologist to model and predict these hydrological and environmental systems more accurately and reliably, especially at large scales. Soil moisture is however a challenging quantity to be measured. It is difficult to define as it may depend on the context and the observation platform, and is sometimes assessed indirectly as being the difference in balancing other terms like precipitation, runoff and evaporation. In situ soil moisture measurements are continuous, accurate and direct but have limited spatial coverage and a high operational cost. They are also difficult to scale up given the substantial local spatial variability. Similarly, airborne observations have limited spatial and temporal coverage, and a high operational cost. Spaceborne satellite observations have wide spatial coverage, are becoming operationally accessible, but are an indirect measure of soil moisture. They are available at different spatial resolutions, and subject to temporal gaps and limitations in some regions. The provision of ongoing reliable daily soil moisture products at continental scale is a challenge task in the face of these features. There are several methods for characterising uncertainty/error of satellite retrievals in order to evaluate and blend them. Most of these methods rely on validation of the retrievals with in situ or airborne observations, which makes it difficult to evaluate satellite soil moisture products at large scales. In addition, uncertainty estimation is complicated by representativeness and scaling errors. These methods normally do not estimate the uncertainty of each data set but are limited to estimating difference between them. For example, the triple collocation methodology has recently been proposed to characterise error structure of remotely sensed observations and to blend them together into a better product. It still cannot handle the temporal gap issue and its performance depends heavily on the selection of a "reference" product. In order to characterise uncertainty/error structure of satellite data sets, and create an ongoing reliable daily soil moisture product, we focus on a so-called double collocation methodology. It requires only two independent time series of collocated soil moisture observations, although it can be easily extended to handle three or more sources of observations. After exploring temporal structure of in situ observations, linear state space models are introduced to estimate the error structure, the underlying soil moisture status, and mapping functions between observations and the underlying status simultaneously. Unlike existing techniques, it does not require the assumption of a reference data set as 'ground truth'. The temporal gaps are smoothed out via making use of all the satellite observations as well as imposing some appropriate temporal structure in the underlying soil moisture status time series. The proposed methodology is preliminarily demonstrated on soil moisture observations in the Murrumbidgee catchment region, Australia. The observations are retrieved from the Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) and the Advanced Synthetic Aperture Radar (ASAR) global monitoring spaceborne sensors. The paper provides discussion on further development and evaluation of this methodology for creating a reliable daily continental soil moisture product.