Assessment of Errors in AMSR-E Derived Soil Moisture
Soil moisture derived from passive microwave satellites provides information at a coarse spatial scale, but with temporally frequent, global coverage that can be used for monitoring applications over agricultural regions. Passive microwave satellites measure surface brightness temperature, which is largely a function of vegetation water content (which is directly related to the vegetation optical depth), surface temperature and surface soil moisture at low frequencies. Retrieval algorithms for global soil moisture data sets by necessity require limited site-specific information to derive these parameters, and as such may show variations in local accuracy. The objective of this study is to examine the errors in passive microwave soil moisture data over agricultural sites in Canada to provide guidelines on data quality assessment for using these data sets in monitoring applications. Global gridded soil moisture was acquired from the AMSR-E satellite using the Land Parameter Retrieval Model, LPRM (Owe et al., 2008). The LPRM model derives surface soil moisture through an iterative optimization procedure using a polarization difference index to estimate vegetation optical depth and surface dielectric constant using frequencies at 6.9 and 10.7 GHz. The LPRM model requires no a-priori information on surface conditions, but retrieval errors are expected to increase as the amount of open water and dense vegetation within each pixel increases (Owe et al., 2008) Satellite-derived LPRM soil moisture values were used to assess changes in soil moisture retrieval accuracy over the 2007 growing season for a largely agricultural site near Guelph (Ontario), Canada. Accuracy was determined by validating LPRM soil moisture against a network of 16 in-situ monitoring sites distributed at the pixel scale for AMSR-E. Changes in squared error, and pairwise correlation coefficient between satellite and in-situ surface soil moisture were assessed against changes in satellite orbit and satellite derived vegetation density from the Normalized Difference Vegetation Index (NDVI). Results indicate that daily LPRM soil moisture values contain significant noise. Errors in soil moisture were reduced when smoothing and weekly averaging were used on the LPRM data set. Errors in satellite-derived soil moisture were lower during spring when soil was generally bare, and higher during periods of dense agricultural growth, indicating that the retrieval of soil moisture under the biomass of the crops found in this region is not reliable. Overall, the LPRM provides a good coarse-resolution estimate of surface soil moisture during low biomass conditions which can be used for monitoring applications in Canada.
Space Geodesy to Improve the Hydrology Modelling at the Global/Regional Scales
The space gravity mission GRACE is now orbiting the Earth for more than 6 years, providing monthly data about mass distribution inside the Earth system. The mission was primarily designed to monitor the climate system, and in particular to improve the hydrology modelling by providing high quality data with a global coverage. We now have enough data to study the inter-annual signal, and analyse the discrepancies between the GRACE derived mass distribution and what is expected from global hydrology models as, for instance, the Land Dynamic model. Focusing on areas where the inter-annual hydrology signal is large and spatially consistent, we investigate the mass distribution fluctuations both in GRACE data, in land hydrology from altimetry and in hydrology models, in order to understand the differences, as a first step to model improvement.
Surface Soil Moisture Distributions at the Field Scale in the Southern Great Plains
Surface soil moisture variability plays a leading role in land surface-atmosphere interactions. Its variability and distribution across a variety of scales has a significant impact on how those interactions are observed and modeled. Combining data from several field experiments in Oklahoma and Texas, fine scale to field scale variability is characterized and evaluated as power law distributions. The Cloud Land Surface Interaction Campaign (CLASIC) in Oklahoma 2007, and the Bushland Evapotranspiration and Agricultural Remote Sensing Experiment (BEAREX) in Texas in 2008 addressed a large number of science questions pertaining to land surface atmosphere interactions. Individualized experimentation on small scale variability at the 5 cm to 200 m scale were conducted using calibrated dielectric soil probes, confirming results from previous experimentation in this region (Washita '92 in Oklahoma) that soil moisture has a power law distribution across a wide range of scales. There were four primary land cover types investigated: dryland cotton, irrigated cotton, pasture, and harvested winter wheat. Multiple methods of analysis were used to confirm these findings including semivariogram analysis, wavelet analysis, and aggregation analysis.
A Simple Downscaling Algorithm for Remotely Sensed Land Surface Temperature
Assimilation of Multiscale Radiation Products Into a Downwelling Surface Radiation Model
Accurate characterization of total downwelling radiation (i.e., downwelling longwave and downwelling shortwave) reaching the Earth's surface is important for modeling surface hydrological processes. Different satellite-based radiative flux products exist where each has its own spatial and temporal resolutions and error characteristics because each uses different methods and utilizes different remote sensing observations. A data assimilation approach can be used to obtain high-resolution estimates by merging different products with an a priori estimate in a way that extracts the most information while accounting for differences in their error characteristics. In this study, we use two commonly used data assimilation (DA) techniques - the Ensemble Kalman Filter (EnKF) and the Ensemble Kalman Smoother (EnKS) - to assess the effectiveness of generating accurate high-resolution fields conditioned on multiple products. A simple cloud-coupled model forced by a combination of geostationary and polar orbiting remote sensing products provides a high-resolution a priori ensemble estimate. The prior estimate is then conditioned by the Global Energy and Water Cycle Experiment (GEWEX) Shortwave Radiation Budget (SRB) product [Pinker et al., 2003], or International Satellite Cloud Climatology Project (ISCCP) based shortwave flux product [Pinker and Laszlo, 1992], and/or the ISCCP-based longwave flux product [Gupta et al., 1992] using either the EnKF or EnKS routine. A combination of different measurement products with different DA techniques is investigated to assess DA effectiveness. When compared against ground-based measurements, preliminary results suggest a multiscale DA approach improves radiative flux estimates, effectively downscaling the relatively coarse (in space and time) measurements while simultaneously reducing the amount of uncertainty across that ensemble. Analysis shows that the covariance structure between longwave and shortwave fluxes is limited especially in the absence of large-scale cloud systems. Furthermore, a comparison between the EnKF and EnKS performance suggests covariance structure is bound within a finite temporal window. Beyond this temporal window spurious correlations may exist. An additional advantage of the EnKS over that of the EnKF is that information from daytime shortwave fluxes can be transferred to nighttime longwave fluxes, but that this information transfer is limited due to a weak covariance structure.
How much improvement can precipitation data fusion achieve based on the Multiscale Kalman Filter framework?
Precipitation data is one of the most important inputs for land surface model simulations. With the advancement in measuring techniques and the improvement in model simulations, more and more precipitation data products become available. These precipitation data are often associated with different spatial resolutions and accuracies. Therefore, there is an increasing need of producing fused precipitation products that take advantage of the strengths of each individual precipitation data product. In this study, we use the Mulitscale Kalman Filter (MKF) framework to fuse individual precipitation products that have different spatial resolutions with different accuracies, and explore the amount of improvement in precipitation that can be achieved through the MKF framework. Two types of errors associated with precipitation, i.e., white noise and biased noise, are investigated. We use the spatial correlation and spatial root mean square error to evaluate the improvements in precipitation accuracy. Our study shows that: (1) the MKF fused precipitation can significantly recover the loss of spatial pattern in precipitation due to the uncertainties associated with the white noise at coarser scales, while the recovery on the precipitation due to this type of uncertainty at the finer scales is much smaller; and (2) there is only limited improvement in precipitation at any scale using the MKF precipitation fusion approach on the biased type of errors.
A Rapid, Object-Oriented Approach to Regional-Scale Wetland Mapping in the Congo River Basin
Wetlands continue to be lost and degraded faster than any other ecosystem despite their importance to hydrology, biogeochemistry and biodiversity. A rapid, regional-scale method for creating wetland inventories is urgently needed to improve hydrological and biogeochemical models as well as conservation efforts, especially for threatened areas that lack baseline information. The wetlands of the Congo River Basin are among the world's largest but are understudied and face imminent development pressures. The potential of combining radar imagery (JERS-1/SAR GRFM) with elevation data (SRTM-derived DEM and derivatives) using object-oriented analysis to rapidly map Congo wetlands is investigated in this study. Findings are comparable to previous, sub-regional wetland analyses of the basin. An accuracy assessment will be presented that compares the results to a partial, reclassified reference map of the basin (FAO Africover). The method shows promise for deriving basic wetland inventories and monitoring programs for the Congo River Basin and other large tropical riverine environments.