A ground reference site for detailed studies of remote sensing of soil moisture
The authors discuss establishment of an experimental site for studies of passive microwave remote sensing technologies and their applicability to observe components of the water cycle. Despite their apparent usefulness, the quantitative aspects of these observations are not well known. There is wide consensus within the scientific and engineering communities that validation of these technologies is an important challenge. To improve the quantitative value of remotely-sensed observations of the water cycle the authors plan developing and using innovative validation techniques. The cornerstone of these efforts is development of a small (~1 km2) prototype experimental validation site. The site will be extensively instrumented with both in-situ and remote sensors so that the complete water cycle and important ancillary data can be carefully characterized at several spatial scales for long periods of time. Initially the focus will be on validating remotely-sensed observations of soil moisture. The site will be fully characterized with respect to topography, soil types, and vegetation. The authors will collect detailed precipitation data using a cluster of rain gauges and NEXRAD information. Atmospheric data such as air temperature, humidity, pressure, wind direction and velocity, and solar radiation will be provides by sensors placed on two towers within the site. These will include eddy covariance evapotranspiration observing systems. Soil moisture and soil temperature vertical profile data will be collected at numerous points of several clusters of wirelessly connected small-scale networks using time-domain reflectometry. These fixed measurements will be complemented by measurements of greater spatial extent made periodically with soil moisture impedance probes. Vegetation measurements will be provided on a systematic basis. Remote sensing data will be provided by a dual-polarized L-band microwave radiometer on both a tower and a mobile platform. Evaporation, dew, and rainfall interception are other variables that will be addressed through targeted campaigns. During those campaigns additional instrumentation such as scanning lidar for observations of water vapor concentration above the canopy level will be deployed. This site will be a community resource: the data generated at the site will be available instantly to other researchers through the use of wireless technologies and the world wide web, and the community is encouraged to co-locate other instruments. The site will be operated by Iowa State University and University of Iowa staff and colleagues at the USDA ARS National Soil Tilth Laboratory. The authors discuss the specifics of the site design and instrumentation as well as the schedule for the site deployment.
Evaluation of Uncertainty in Nested Flood Forecasts by Coupling a Multifractal Precipitation Downscaling Model and a Fully-Distributed Hydrological Model
Despite progress in satellite remote sensing, the use of space-borne precipitation fields in hydrological models remains elusive, in particular with respect to the space and time scales appropriate for flood forecasting. A major limitation towards use of coarse satellite products is the lack of a formal framework for downscaling precipitation fields in space and time over the catchment of interest. In this work, we develop and test a hydrometeorological forecasting procedure intended to evaluate the uncertainty incorporated into streamflow predictions through the use of downscaled precipitation products. Our approach relies on using a space-time multifractal model to downscale a coarse precipitation product, such as a satellite observation, and generating an ensemble of precipitation fields at high resolution. These synthetic fields are used to force a fully distributed hydrological model known as the TIN-based Real-time Integrated Basin Simulator (tRIBS) for flood prediction at multiple, nested locations. For this study, we first investigate the scaling properties of precipitation derived from the NEXRAD radar network in the Arkansas Red River Basin for the 1997-2003 summer months. We then calibrate a multifractal model based on a log-Poisson generator, exploring the linkages between the model parameters and the large scale meteorological observables. We also evaluate the accuracy of the downscaling products relative to the high resolution observed fields. Subsequently, we force tRIBS model with the synthetic downscaled precipitation ensemble to predict streamflow in the Baron Fork basin in Oklahoma and at nested interior locations. The resulting ensembles of synthetic hourly hydrographs and observed streamflow values are then post-processed to verify the forecast procedure. Scalar and non-scalar measures are used to evaluate the forecast reliability, resolution, sharpness and bias. The plausibility of the consistency condition for the ensemble forecast procedure is also investigated through the analysis of the verification rank histogram. Furthermore, we identify the catchment scale dependency in the forecast performance through analysis at multiple nested basins. To conclude, we summarize how ensemble forecast metrics can be used to quantify propagation of uncertainty from downscaled precipitation products to distributed flood forecasts.
The Potential for Application of a Physically-based Snow Model in Hydrologic Forecasting
There has been a significant push in the hydrologic community towards advancing hydrologic prediction technology. Efforts to improve streamflow prediction can be focused in any number of ways, including the introduction of more complex, physically-based models into operational forecasting systems. The National Weather Service (NWS) SNOW17, a conceptually-based snow accumulation and ablation model, has been used relatively unchanged for streamflow prediction for several decades. In a series of studies, the potential for improving the accuracy of streamflow predictions through the use of an energy balance snow model was investigated. In this work, the Snow-Atmosphere-Soil Transfer (SAST) model, which uses the energy-balance method, is evaluated against the SNOW17 model for the simulation of seasonal snowpack and the prediction of ensembles of future snow water equivalent. The study is structured around the current NWS forecasting system, assuming these are the constraints in which an advanced operational snow model will be required to work in the near term. Results show that improvements to snowpack predictions (and subsequent streamflow predictions) through the use of an energy balance snow model will be minimal, given the current forecast system and available data. The impact of input uncertainties on model results is also highlighted.
L-Band Microwave Observations over Land Surface using Two-Dimensional Synthetic Aperture Radiometer
A number of studies have demonstrated the potential capabilities of passive microwave remote sensing at L- band (1.4 GHz) to measure surface soil moisture. Aperture synthesis is a technology for obtaining high spatial resolution at long wavelengths with a practical radiometer antenna. During the Soil Moisture Experiment in 2003 (SMEX03), the Two-Dimensional Synthetic Aperture Radiometer (2D-STAR) was flown onboard a NASA P3B aircraft over different climate regions with various vegetation conditions in Alabama, Georgia, and Oklahoma. 2D- STAR is one of the first airborne synthetic aperture radiometers capable of producing multiple polarized multi- angular brightness temperature data over land. In this study, brightness temperature data collected during SMEX03 is presented and analyzed for various land cover types, which include pasture, crop field, and forest. The sensitivity of the brightness temperature to soil moisture at forest sites is of particular interest in understanding its potential to estimate soil moisture under dense vegetation cover. Advantages of utilizing dual polarized multi- angular L-band data and implications for processing L-band interferometric data from space by the future Soil Moisture and Salinity Mission (SMOS) are discussed.
The Determination of Soil Hydraulic Properties from Remotely Sensed Surface Temperature in Land Surface Models
Soil Hydraulic Properties (SHPs) play an important role in regulating heat and moisture fluxes in land surface models.ï¿½ However, studies have shown that the common method of determining SHPs from soil texture class is inadequate (Gutmann and Small 2005).ï¿½ Because of this, a new method of determining SHPs is required.ï¿½ We present an analysis of the determination of SHPs from surface temperature (Ts).ï¿½ Ts is related to SHPs because soil moisture controls evaporative cooling of the surface as well as thermal conductivity and specific heat of the soil.ï¿½ Ts datasets are available globally from numerous remote sensing platforms, and thus the ability to determine SHPs from Ts will greatly improve land surface models used in climate and weather forecasting. ï¿½ We use the Noah land surface model to investigate the sensitivity of Ts to SHPs in semi-arid environments from central New Mexico, Kansas, and Oklahoma.ï¿½ To do this, we first calibrate one parameter related to the aerodynamic roughness of the surface using Ts during hot, dry periods, when Ts will be insensitive to SHPs.ï¿½ Next we use drydown periods following rainstorms to calibrate SHPs based on measured Ts.ï¿½ To verify the results, we calibrate SHPs with measured latent heat flux (LH) at the same sites.ï¿½ Our results indicate that SHPs determined from Ts are comparable to those determined from LH. However, when SHPs are determined separately for separate time periods at the same site, there is greater variation in the SHPs determined from Ts than in those determined from LH. Finally we show that SHPs derived from remotely sensed Ts are similar to those derived from field measured Ts and LH. ï¿½
Soil Moisture Retrieval Uncertainty From Soil and Vegetation Heterogeneity Over a Topographic Surface
At the interface between the land surface and atmosphere, soil moisture governs evapotranspiration, infiltration and runoff processes. Current knowledge of the soil moisture is poor, although a satellite L-band passive microwave mission is planned that will monitor the global surface soil moisture. This Soil Moisture and Ocean Salinity (SMOS) mission uses synthetic aperture techniques to achieve a resolution of 35-50km, and is scheduled for launch in September 2007. Besides noise in the measurements, determination of the soil moisture from the brightness temperature measured by SMOS has a number of sources of error in: physics of the forward model, parameters in the model and from surface heterogeneity in the field-of-view. In fact, to achieve a mission specification of 4% volumetric soil moisture accuracy, the retrieval of soil moisture is more sensitive to surface temperature, open water coverage and melting snow than can reasonably estimated from available data. Therefore, a data assimilation framework is needed to determine soil moisture from passive microwave measurements. Here, we examine soil moisture retrieval error from topographic effects. Local slope affects the pathlength through vegetation and the soil emissivity given by the Fresnel equations. Previous work has shown that the mean error is less than 0.6% for a triangular hillslope with a maximum coverage of 30% slopes at 30ï¿½ under the assumption that the soil and vegetation is uniform across the scene. For a field-of-view as large as that of SMOS, an assumption of soil and vegetation homogeneity is not realistic. Variation in these parameters may occur with elevation, aspect and location. This paper quantifies soil moisture retrieval error for simple topographic scenarios, and tests the results for a DEM transect. The forward model aggregates brightness temperatures over a scene for both horizontal and vertical polarization at a range of look angles. In the forward model, soil moisture, surface temperature and vegetation optical depth were assumed to be a simple function of either slope or aspect. An inverse model was used to retrieve soil moisture based on the minimization of a cost function. The soil moisture retrieval errors were compared to those expected from a topographically flat scene with comparable parameter heterogeneity. These errors must be taken into account within the overall error budget of the SMOS or similar missions.
Remotely Sensed Potential Evaporation Estimates for Hydrologic Modeling
This study explores a methodology solely dependent on remote sensing information to capture both the current climate signal and the spatial variability of daily potential evaporation (PE) by taking advantage of the new generation of Earth Observation satellites (i.e., MODIS sensor). PE, a required input for most hydrologic models, is typically obtained from pan evaporation estimates, or in some cases, from ground-based meteorological measurements at limited point locations. We focus our efforts on development of a ï¿½ï¿½ï¿½ï¿½stand-aloneï¿½ï¿½ï¿½ï¿½ method to derive daily estimates of PE without the need for ground-based observations. The procedure is based on the Priestley-Taylor equation, incorporating a previously developed daily net radiation model during cloudless days. We then apply a simple algorithm using theoretical clear-sky net radiation and potential evaporation (linearly interpolated values during clear days), along with a daily cloud fraction to estimate net radiation and potential evaporation under cloudy conditions. For initial validation, point scale comparisons are undertaken using the single pixel value from MODIS corresponding to four ground-based observation sites covering a range of hydroclimatic conditions and biomes: Bondville (IL), Goodwin Creek (MS), Audubon (AZ) and Westville (OK). Preliminary results over a several year period (2001-2004) at three of the sites (Bondville, Goodwin Creek and Westville) show good correlation (R=0.875) and bias (0.227mm/day) at the daily time step. Results are further improved when aggregated to the monthly timescale (R=0.953, bias=0.197 mm/day). Performance at the Audubon site (semi-arid biome) is less satisfactory (R=0.820 and bias=2.025 mm/day at the daily time step). However, results are extremely promising and show the potential for application to hydrologic modeling and water-balance studies in both gauged and un-gauged basins. Further work is on-going to investigate deficiencies in semi-arid regions and to improve estimates under cloudy conditions.
Multi-scale water balance evaluation using remote sensing and agro/eco-hydrological modeling
A multi-step approach is applied to upscale the water balance from field to catchment and to meso-scale region where the meso-scale region is represented by the major island of Denmark, Sjaelland, which has an area extension of 7330 km2. The methodology is based on the use of multi-scale remote sensing data and national GIS data for agro/eco-hydrological modeling using the DaisyGIS model. At field scale (local level), water balance is calculated using local scale parameters representing 3 experimental sites (agricultural land, forest and urban region) and validated using eddy covariance data of latent heat fluxes. The water balance components of the 3 experimental sites varied widely: Evapotranspiration constitutes 67% of precipitation at the agricultural site, 55% of precipitation for an old deciduous forest site and only 42% of precipitation in a dense urban region. Forest evapotranspiration was closely related to leaf area development which was represented using MODIS time series data. At the regional scale, semi-variance analysis of high spatial resolution remote sensing data (Landsat TM) is used to assess the appropriate spatial resolution whereby local scale model parameters can be adapted for deterministic hydrological modelling. Despite a high sensitivity of water balance calculations to land cover and soil types, catchment simulations based on 500 m grid representation of land cover and drainage patterns were statistically similar to catchment simulations using 30 m grid representation. During low-flow conditions, a high agreement between catchment simulations and discharge data was observed for 30 catchments. Consideration of urban runoff improved simulations in the capital region. Furthermore, the soil water percolation of forest regions were found to be very sensitive to model paramterization. During winter, the percolation of soil water to groundwater was comparable to larger-scale trends in groundwater level dynamics which had been evaluated for 10 major climate zones at Sjaelland.