Trend-preserving blending of passive and active microwave soil moisture retrievals

A series of satellite-based passive and active microwave instruments provide soil moisture retrievals spanning altogether more than three decades. This offers the opportunity to generate a combined product that incorporates the advantages of both microwave techniques and spans the observation period starting 1979. However, there are several challenges in developing such a dataset, e.g., differences in instrument specifications result in different absolute soil moisture values, the global passive and active microwave retrieval methods produce conceptually different quantities, and products vary in their relative performances depending on vegetation density. This paper presents an approach for combining four passive microwave products from the VU University Amsterdam/National Aeronautics and Space Administration and two active microwave products from the Vienna University of Technology. First, passive microwave soil moisture retrievals from the Scanning Multichannel Microwave Radiometer (SMMR), the Special Sensor Microwave Imager (SSM/I), and the Tropical Rainfall Measuring Mission microwave imager (TMI) instruments were scaled to the climatology of the Advanced Microwave Scanning Radiometer — Earth Observing System (AMSR-E) derived product and then all four were combined into a single merged passive microwave product. Second, active microwave soil moisture estimates from the European Remote Sensing (ERS) Scatterometer instrument were scaled to the climatology of the Advanced Scatterometer (ASCAT) derived estimates. Both were combined into a merged active microwave product. Finally, the two merged products were rescaled to a common globally available reference soil moisture dataset provided by a land surface model (GLDAS-1-Noah) and then blended into a single passive/active product. Blending of the active and passive data sets was based on their respective sensitivity to vegetation density. While this three step approach imposes the absolute values of the land surface model dataset to the final product, it preserves the relative dynamics (e.g., seasonality and inter-annual variations) of the original satellite derived retrievals. More importantly, the long term changes evident in the original soil moisture products were also preserved. The method presented in this paper allows the long term product to be extended with data from other current and future operational satellites. The multi-decadal blended dataset is expected to enhance our basic understanding of soil moisture in the water, energy and carbon cycles.