Evaluating and Improving the Information Content of Satellite Soil Moisture Measurements

Soil moisture plays an important role in controlling the exchanges of water and energy at the land-atmosphere interface and profoundly influences the spatial and temporal variability of weather and climatic conditions. As a result, many satellite missions have focused on providing space-borne measurements of soil moisture, primarily from passive microwave radiometry. The recently launched NASA Soil Moisture Active Passive mission is the latest among a series of similar missions to measure soil moisture from space. Compared to the ground measurements, the stochastic noise in remote sensing measurements tends to be large. In addition, the remotely sensed soil moisture measurements generally have large biases and moderate to fair correlations when compared to in-situ measurements, due to representativeness issues from sensing depth differences, instrument errors and retrieval algorithm errors. Normalization methods to match statistical moments against reference data are usually used to circumvent systematic errors in these retrievals.  Studies, however, have shown that such methods are not effective in improving correlation skills of the satellite data due to the large fluctuations at short time scales.

In this presentation, we examine the use of a suite of methods for evaluating and improving the information content of satellite soil moisture retrievals. While accuracy based performance measures are often used to evaluate the satellite soil moisture measurements, they are insufficient to provide estimates of performance for unobserved scenarios or when reliable reference datasets do not exist. In such instances, the use of information theory based measures that provide estimates of complexity and information content of the soil moisture signal can be an effective way to quantitatively evaluate these measurements. In this work, we focus on the application of both accuracy-based methods (comparison to ground measurements, triple-colocation methods) and information theory based approaches for evaluating satellite soil moisture measurements. In addition, we will also demonstrate and evaluate the use of spectral methods to reduce the stochastic error components in the satellite retrievals. The evaluations will be performed for soil moisture retrievals from recent missions such as SMAP, SMOS, ASCAT an AMSR2.