Towards land surface model validation from using satellite retrieved soil moisture

Abstract

Land surface model validation at distributed scales is important for model improvements. Recent advances in satellite technology provide an opportunity for distributed calibration and validation of land surface models. In the past years, a number of active and passive microwave soil moisture products have become available. While passive microwave soil moisture is the preferred approach for soil moisture observation, its disadvantage is the coarse spatial resolution it affords. Moreover, many of the available satellites use sub-optimal wavelengths, and the satellite retrieval algorithms are still under development. Consequently, the accuracy of these satellite data sets needs to be verified prior to their application. However, the spatial and temporal discrepancies between in-situ monitoring and satellite footprint retrievals continue to make absolute verification of satellite retrieved soil moisture a difficult problem. The Advanced Microwave Scanning Radiometer-2 (AMSR-2) onboard the Global Change Observation Mission 1 - Water (GCOM-W1) was launched by JAXA in May 2012. AMSR-2 is a follow on of the AMSR-Earth Observing System (AMSR-E) onboard Aqua and of the AMSR onboard the Advanced Earth Observing Satellite 2 (ADEOS-II). By combining data from AMSR, AMSR-E and AMSR-2, a 20-year record of near-continuous C-band measurements of soil moisture content is expected to be available, starting from 2001. This study makes an inter-comparison between in-situ data from the OzNet soil moisture network (www.oznet.org.au), the AMSR-2 soil moisture product, and simulated soil moisture using JULES (Joint UK Land Environment Simulator) for the period July to December 2012. The area selected is a 60 km × 60 km study site in Yanco, NSW, Australia (34.561°S, 35.170°S, 145.826°E, 146.439°E). 10 km and 25 km soil moisture products from the descending orbit of AMSR-2, which has a repeat time of 1 to 2 days, has been used. The JULES land surface model was run at hourly time-steps and approximately 1 km (0.01°) resolution for the entire 60 km × 60 km Yanco area, which coincides with twenty-five 10 km and four 25 km AMSR-2 product grids at hourly time-steps. Due to the co-location between in-situ monitoring stations and AMSR-2 grids, comparison between both data sets was only possible at five 10 km and two 25 km AMSR-2 pixels. Where in-situ stations are available, time series of AMSR-2 soil moisture and JULES simulations were validated against in-situ measurements. AMSR-2 products and JULES simulations were also compared against each other. The average RMSD for both 10 km and 25 km products were found to be 0.05 m 3 /m 3 when compared to in- situ data, which meets the target accuracy of the mission. The AMSR-2 soil moisture was used to evaluate simulated soil moisture. Being a consistent product across time and space, AMSR-2 soil moisture can be used to identify where model simulations are inaccurate due to forcing data, parameter assignment or model physics. Whilst the opportunity in using AMSR-2 soil moisture to validate land surface models run at distributed scales was demonstrated, this study could not conclude whether the satellite or simulated soil moisture is more accurate due to possible inaccuracies in the current radiative transfer model, parameterization of soil and vegetation characteristics and prescription of precipitation data in the land surface model. The study also indicated prospects in further studies for better understanding of the Yanco site in relation to 1) representativeness of the sites used for validation and 2) effects caused by vegetation and standing water within the satellite footprint to improve the retrieval algorithm of AMSR-2 soil moisture for Australian conditions.