Incorporating the Land Data Assimilation System into Water Resource Management and Decision Support Systems

Kristi R. Arsenault, Paul R. Houser, and David A. Matthews

The Land Data Assimilation System (LDAS) team at NASA’s Goddard Space Flight Center (GSFC) is developing a system to aid water resource managers in making flood and drought assessments and predictions. This system runs multiple land surface models, assimilating and using the latest surface observations and remotely sensed data as both operational and retrospective forcings. The emphasis of LDAS is on capturing the most realistic representations of land surface states and dynamics over large areas and at high resolutions. The main hydrometeorological variables on which LDAS focuses are soil moisture, evaporation, snow cover, runoff, precipitation, and surface energy budget variables.
The United States Bureau of Reclamation (Reclamation) relies heavily on accurate and timely hydrometeorological information for river basin management. More than 80 percent of the water supply in the West is provided by snowpack runoff, so to generate accurate flood forecasts, Reclamation must have accurate estimates of snow water equivalent and must be able to monitor the evolution of the snowpack into runoff. In addition to generating flood forecasts, Reclamation also requires the ability to predict the effects of drought conditions on agricultural production, on the public, and on wildfire vulnerability. To address these important issues, output from LDAS Land Surface Models (LSMs) can be integrated into the river basin decision support systems that Reclamation’s managers routinely use to manage western river systems. Ultimately, using the LDAS LSM will improve overall flood and drought risk analysis and prediction.

LDAS Background
LDAS research efforts are divided between two main projects: North American LDAS (NLDAS) and Global LDAS (GLDAS). Both projects share the same underlying objectives: LSM development, use of observations as forcing data, and assimilation of different satellite and radar information. A few differences do exist between the two projects, but as the projects develop side-by-side they strengthen each other’s growing utility for applications and atmospheric prediction models.
Overall, the NLDAS and GLDAS projects are characterized as real-time, distributed, uncoupled, land-surface simulation systems on a U.S. national domain at .125-degree resolution and on a global domain at .25-degree resolution. Both LDAS projects use a suite of different LSMs running in tandem on these grid systems and driven by common surface forcing.
Land Surface Models currently incorporated into LDAS include Mosaic, the National Centers for Environmental Prediction model, the Oregon State University model, the United States Air Force model, the Office of Hydrology model, and the Community Land Model. Additional LSMs being brought into LDAS include the Variable Infiltration Capacity model and the Catchment Land Surface Model.
Other major components of the LDAS projects include replacement of atmospheric model-based forcing with observations, assimilation of remotely sensed and in-situ measurements into the LSMs, and output validation and calibration of the LSMs. By using observations to drive land surface models, such as precipitation and radiation, biases present in coupled atmospheric-land surface model systems can be avoided.
In the NLDAS project, the modeled precipitation is replaced with a merged product consisting of Stage IV Weather Service Radar-88 Doppler, gauge precipitation, and modeled precipitation from the National Centers for Environmental Protection atmospheric Eta-based 4-D Data Assimilation System. Downward shortwave and longwave radiation products used are from GOES satellites. GLDAS also has replaced some of its modeled fields with satellite-derived precipitation, such as the NASA GSFC 3-hourly merged-satellite product, which includes geostationary infrared, Special Sensor Microwave/Imager, and Tropical Rainfall Measuring Mission data. As for radiation fields, GLDAS uses global downward shortwave and longwave radiation products from the Air Force Weather Agency.
LSMs are run over both LDAS domains using vegetation-based “tiles” to simulate variability below the scale of the model grid squares, with each tile representing an area covered by a given vegetation type. Both NLDAS and GLDAS use the 1-km University of Maryland vegetation classification scheme, which is based on the climatology of AVHRR remote sensing data. A near-real-time, satellite-based, leaf-area index derived from AVHRR and MODIS data is currently being used in some of the LSMs.

Incorporation of Data
Assimilation Techniques
Substantial surface forcing errors in coupled land-atmosphere, four-dimensional, data assimilation systems are a major reason why LDAS and data assimilation techniques were developed. In a real-time context, data assimilation can provide quality assurance and validation of observations and can provide rapid identification and diagnosis of barely perceptible problems. Several satellite products to be used in LDAS include snow cover from MODIS and soil moisture from the Aqua satellite’s Advanced Microwave Scanning Radiometer. Assimilating these different satellite data into LDAS will produce a more accurate description of the Earth system than can be provided by the satellite observations driving the land model simulations alone. Such descriptions would benefit estimates of snow depth and cover in the winter and spring months needed to forecast amounts that could contribute to river runoff, especially in river basin areas like the Columbia River that are heavily affected by spring snow melt. Assimilation of snow information can directly improve the day-to-day snow states in the LSM and can act as a means of diagnosing problems in LSM physics. The improved LSM, unassisted by observational data, can then provide improved snow simulations.

Integration of LDAS LSM Output into Reclamation’s Decision Support Systems
Demonstrating the use and added value of LDAS output in Reclamation’s operations and decision-support tools is becoming a major part of LDAS’s applications research. Reclamation applies the latest emerging technologies in decision support modeling to enable river systems operations managers to make better-informed decisions. The key river basin modeling technology includes RiverWare and the Agricultural Water Resources Decision Support — Evapotranspiration Toolbox system (AWARDS-ET Toolbox). RiverWare is a river modeling and water accounting system that provides a flexible framework for developing and running site-specific models that incorporate the “law of the river” and other policy constraints. RiverWare simulates the routing of the river flow operations through dams and hydropower plants and maintains water delivery contracts to irrigators and to recreational, municipal, and industrial users. Even though Reclamation uses several hydrological modeling tools for RiverWare, such spatially-distributed fields as soil moisture are not fully accounted for in their operations. Including such fields in their operations could help increase the accuracy of Reclamation’s water resource accounting.
Determining the value added by LDAS output into RiverWare and ET Toolbox will require an examination at various spatial and temporal scales of the different vegetation and soil parameters being used for LDAS and for running the LSMs. This examination may improve water use estimates, especially during prolonged drought. The ET Toolbox estimates the daily surface water use requirements of riparian and crop vegetation and of open water evaporation estimates. Integrating different LSM output into the ET Toolbox and transferring the LSM parameterizations and physics to different resolutions should improve ET drought estimates. A plan to estimate how soil moisture fields can improve ET estimates and the efficiency of water management is another major goal.
Two means of collaboration between LDAS and Reclamation include (1) the exchange of data to help evaluate the ability and suitability of the current LDAS to fit into Reclamation’s water management operations and decision support tools, and (2) mutual validation efforts that will improve the parameterization and physics of LDAS LSMs and that will incorporate scaling issues to render improved output specific to Reclamation’s needs. NASA recently funded an LDAS extension project called Land Information Systems (LIS). LIS is designed to run outlined LSMs at 1 km resolution and to be used as a significant high-end computing applications project, which would provide Reclamation even higher resolution output to facilitate their need for information at the catchment levels.

Acknowledgments
Funding for the GLDAS project is provided through the NASA Earth Science Enterprise Program. The NLDAS project has been funded primarily through the NASA Land Surface Hydrology Program. Special thanks are expressed toward Brian Cosgrove, Jon Gottshalck, and Dr. Xiwu Zhan for their comments and feedback on this article and to the River Systems and Meteorology Team at Reclamation for providing additional information. Also, we gratefully acknowledge all those who have contributed to the overall LDAS projects from NASA GSFC, other agencies, and universities.

About the Authors
Kristi Arsenault is a research analyst with the University of Maryland, Baltimore County, Goddard Earth Science and Technology Center, and she conducts her research at NASA GSFC. Ms. Arsenault has been involved for almost two years with the GLDAS project and investigates ways in which both the NLDAS and GLDAS projects could be applied to different societal issues, such as water resources management. Ms. Arsenault can be contacted at kristi @hsb.gsfc.nasa.gov.
Dr. Paul Houser is in charge of the Hydrological Sciences Branch at NASA GSFC. Dr. Houser is a leading investigator on the two LDAS projects, and his research interests include local to global land surface-atmospheric observation (both in-situ and remotely sensed) and numerical simulation, development and application of hydrologic data assimilation methods, and multi-scale moisture investigations. He may be contacted at houser@hsb. gsfc.nasa.gov.
Dr. David A. Matthews is the manager of the River Systems and Meteorology Group at the Bureau of Reclamation. Dr. Matthews’ group works on the Watershed and River Systems Management Program, NEXRAD algorithm development and testing for snow water equivalent estimates, agricultural water conservation decision support system development for water supply and demand analysis and forecasting, and early warning system applications of NEXRAD data for dam safety. Dr. Matthews can be contacted at [email protected].

Back