Retrieval of soil surface roughness from active and passive microwave observations

Spatial and temporal variation in soil moisture plays a significant role in establishing efficient irrigation scheduling, climate change prediction, and sustainable land and water management. Passive microwave remote sensing at L-band is widely recognised as the preferred technique to measure surface soil moisture globally, with spatial resolution ranging from 40-100km. However, passive microwave soil moisture retrieval is highly dependent on ancillary data such as surface roughness, which is difficult to characterise by ground measurement. The National Aeronautics and Space Administration (NASA) is developing the Soil Moisture Active Passive (SMAP) mission, scheduled for launch at the end of October 2014. SMAP will deploy both L-band active (radar) and passive (radiometer) microwave instruments to enhance the soil moisture retrieval capabilities. Consequently, deriving roughness information from the active microwave sensor onboard SMAP provides an opportunity to solve the aforementioned problem. However, it is unclear if roughness parameters derived from active microwave data can be used directly in passive microwave retrievals. This paper presents a series of roughness-related analysis using data from the Soil Moisture Active Passive Experiment (SMAPEx), conducted in south-eastern Australia in 2010-2011. The SMAPEx airborne instrumentation comprised an L-band radar (PLIS) and radiometer (PLMR), replicating the active/passive configuration of the SMAP satellite. Intensive ground soil moisture sampling was undertaken concurrently with each flight, and surface roughness was sampled within each major land cover type. In this study, three 1km pixels with relatively homogeneous bare surface were selected, representing three types of roughness patterns: sinusoidal, flat bench and non-periodic structure. Both soil moisture and roughness parameters of these three pixels were retrieved from PLIS backscatter coefficient and PLMR brightness temperature respectively, and then compared with each other. Results show that soil moisture values retrieved from passive measurements have a much higher accuracy (RMSE=0.05m/m) than those retrieved from active measurements (RMSE=0.11m/m). Active microwave significantly underestimated the soil moisture (Bias=0.76m/m), especially in the top soil moisture range (0.15-0.35m/m). In addition, the studied pixel with non-periodic structure showed the best roughness retrieval accuracy (RMSE=0.35cm) from active measurements compared with ground measurments, while the algorithm tends to overestimate the roughness at the two pixels with periodic structure (RMSE=0.52cm and 0.63cm respectively). Importantly, it was found that the roughness parameter HR retrieved from passive microwave was higher than HR retrieved from active microwave for all pixel types (calculated using two different formulations from retrieved surface root-mean-square height, Bias=0.21 and 0.36 respectively), leading to the conclusion that passive microwave may be more sensitive to larger-scale of roughness than active microwave. However, these results still need to be confirmed with more data points.