APPENDIX D

SUMMARY OF RESULTS FROM GCIP/LSA-E DETAILED DESIGN WORKSHOP

The GCIP/LSA-E Detailed Design Workshop was held in Huntsville, Alabama on 20 - 22 October, 1996 at the Holiday Inn - Research Park. The primary purpose of this workshop was to provide inputs to the design of the overall experiment for the LSA-E during the water years 1998-1999. The Workshop made use of the document entitled "GCIP Studies in the LSA-E - A Discussion Paper" compiled by Dale Quattrochi as a starting point in developing recommended research activities.This document is available through the WEB on the GCIP Home Page or at the URL address: (http://wwwghcc.msfc.nasa.gov/GCIP/). This Appendix contains an abbreviated summary of the results from the work sessions.

The characteristics of the major river basins in the LSA-E are:

Upper Ohio River provides semi-humid, Appalachian headwater signature in Mississippi River hydrograph
Tennessee-Cumberland River provides semi-humid southeast tributary, representative of hydrology in this region.

The features of the Ohio and Tennessee River basins important to the GCIP continental-scale studies include the following:

Topographic effects of the Appalachian Mountains
Heaviest precipitation in the entire Mississippi River basin
Winter-spring precipitation maximum
Winter-spring floods
Synoptic weather systems as major precipitation cause
Some snowmelt effect
Rivers in deep valleys (gulleys)
Dominant contribution to Mississippi River runoff
Few large natural reservoirs, but many manmade [e.g., Tennessee Valley Authority (TVA)]

The features and characteristics listed above led to the emphasis on research studies and modeling for this region to focus on the annual hydrometeorological cycle dynamics and water resources management.

D.1 LSA-E Infrastructure and Related Research

A significant part of the Workshop was a series of presentations on the existing facilities and current research activities in the region which are potentially useful for collecting data needed by GCIP and/or for cooperative research studies with GCIP. A summary of these presentations was given in last year's edition of the GCIP Major Activities Plan (IGPO 1996a).

D.2 Work Sessions

Work Sessions were held in two phases. The first phase addressed three specialized topics while developing an approach to the major research questions on the annual hydrometeorology and water resources that are significant to the success of GCIP. The three topics were:

1. Coupled Hydrologic/Atmospheric Modeling

2. Diagnostic Studies/Energy and Water Budgets

3. Hydrometeorological Prediction and Water Resources Management

The second phase then further developed the specific research and data issues defined during these initial Work Sessions.

GCIP research addresses activities on two scales in each Large Scale Area (LSA). Intermediate-scale area (ISA) activities at spatial scales on the order of 1,000 to 10,000 sq km are phased in with those for each LSA. Small-scale area (SSA) activities at a spatial scale on the order of 100 sq km typically involve efforts requiring intensive observing periods over a concentrated region to study focused issues. The Work Sessions were asked to identify candidate ISA and SSA activities in the LSA-E.

D.3 Coupled Hydrologic/Atmospheric Modeling Work Session

The development and validation of coupled hydrological-atmospheric models is a major scientific objective for GCIP that includes improving the representation of land surface components in models. This Work Session was asked to consider how GCIP can make use of the unique features, infrastructure and data available in the LSA-E to develop and evaluate regional coupled hydrologic/atmospheric models for weather and climate prediction. In particular, it addressed questions such as what coupled modeling issues can be addressed in the LSA-E?; what processes pertaining to characteristics inherent to the LSA-E need to be emphasized?; how can we evaluate the capability of coupled models to simulate the causal mechanisms for interseasonal and interannual variability over the LSA-E?; and what is needed to estimate model parameter values over the annual hydrologic cycle?

The Work Session was also asked to identify the types of data needed for hydrological and atmospheric modeling research; to identify where such data are available in the LSA-E; and to recommend enhancements to assure sufficient data are available for the Water Years 1998 and 1999.

The coupled hydrologic-atmospheric modeling Work Session recommended research tasks in four areas and summarized in the remainder of this section.

D.3.1 Model Grids and Coordinate Systems

The current status of the three regional models being used by GCIP to provide model output data for budget studies and other applications was reviewed with emphasis on the capability to produce the model output needed during the Water Years 1998 and 1999.

The three regional models producing output for GCIP are archived on a 40 km resolution grid using a Lambert Conformal Map projection true at 100W longitude. However, the "native" grid system resolution varies among the three models. These variations provide an opportunity to investigate the extent to which each of the three regional model grid and coordinate systems are adequate to model the effect of orography on precipitation and the effect of heterogeneous vegetation in the LSA-E.

In addition, these evaluations should include comparisons with higher resolution grids. The Eta model produced model output at 10 km resolution over a portion of the LSA-E during the period of the 1996 Olympics in Atlanta, GA. A model output data set such as this is well suited for comparative evaluation on the effects of grid resolution in capturing orographic effects on precipitation and the effect of heterogeneous vegetation.

D.3.2 Model Initiation

The Work Session considered there is little data available in the LSA-E for coupled hydrologic/atmospheric modeling in both the operational and the research mode. It was recommended that sensitivity studies be conducted on the effects of improved initiation of coupled mesoscale models in very complex regions (such as the LSA-E) with special attention to orography, vegetation, groundwater, and heavily managed runoff.

D.3.3 Modeling Clouds

The Work Session recognized that all aspects of cloud parameterization in atmospheric models could be improved. However, it was recommended that some emphasis should be placed on the problem of representing low-level cumulus clouds. The feedback on the surface energy balance needs to be included in coupled mesoscale models and the parameterization of such clouds evaluated using detailed, satellite based estimates of cloud cover.

D.3.4 Compatibility of Regional and Global Models

It was considered that the relative value of output from regional and global models is largely an open question in the case of LSA-E, and that this may have seasonal characteristics. The Work Session recommended that some priority be given to the evaluation of global model output using regional data sets from the LSA-E. In this regard, it was recommended that GCIP give consideration to the following questions.

(a) Should global model output products be a formal part of the GCIP data base?
(b) Should the model physics be consistent between the regional and global models used at NCEP to produce operational output products?
(c) Is the soil moisture initiation in regional and global models adequate?

D.4 Diagnostic Studies/Energy and Water Budgets Work Session

Determining the time and space variations of the energy and water budgets from daily to seasonal and interannual periods for the continental scale is one of the scientific objectives for GCIP. This Work Session was asked to consider the types of energy and water budget studies that could best be done in the LSA-E that could contribute to the successful achievement of this scientific objective for GCIP. This Work Session was also asked to identify the data requirements needed to conduct energy and water budget studies; to consider how the existing facilities could contribute to these budget studies; and to recommend enhancements to the existing facilities which the GCIP Project should make during the two-year data collection period of Water Years 1998 and 1999.

The Work Session was focused on energy and water budgets and their variations on seasonal to interannual time scales. The primary questions it addressed were:

The Working Group was asked to make specific recommendations with respect to:

(i) Candidate list of small-scale area basins(SSAs)within the LSA-East,

(ii) Candidate intermediate scale area basins(ISAs) within the LSA-East,

(iii) Identification of existing sources to meet data requirements in the LSA-East, and

(iv) Data collection enhancements to existing facilities for the 1998 and 1999 Water Years.

The Group in the Work Session noted that given the overall complexity and heterogeneity of the LSA-E it would be exceedingly difficult to design an observational program that could sample data representative of each micro-climate and ecosystem niche. Thus the group suggested that it would be prudent to suggest the minimum number of SSAs that would sample two major ecosystem types, forests versus cultivated land areas, and regions with distinctive climates, northern versus a southern areas. A survey of existing instrumented sites resulted in recommending that the following sites be considered as candidates for SSA sites:

(1) Goodwin Creek Watershed; Oxford, MS USDA/ARS/NSL

(2) Walker Branch Experimental Watershed; Oak Ridge, TN

(3) North Appalachian Experimental Watershed;Coshocton, OH USDA/ARS

(4) Alabama A&M Experiment Station and Remote Sensing Center; Huntsville, AL

(5) Redstone Arsenal; Huntsville, AL U.S. Army

(6) Panola experimental watershed near Atlanta, GA USGS and NOAA/ERL

The Working Group recommended augmenting or changing locations for the current MOLTS array produced by the coupled mesoscale models to include the candidate SSA sites listed above.

As in all GCIP study areas, precipitation was identified as the most critical variable. It was recommended that the current GCIP mosaic precipitation data set be checked to insure that it was obtaining all of the precipitation networks within the LSA-E. Given the complex terrain and potentially large amounts of data it was suggested that the WSR-88D estimated rainfall would be most useful in conjunction with SSA and ISA study areas.

D.5 Hydrometeorological Prediction and Water Resources Management

The water resources working group focused on how GCIP LSA-E activities could contribute to GCIP's evolving goals with respect to water resources. The group started by identifying some of the most important characteristics of LSA-E with respect to water resources:

1) For water resources purposes, LSA-E consists of the Tennessee-Cumberland and Ohio River systems. The two systems have hydroclimatological similarities, but from a water resource systems standpoint they are much different. The Tennessee River system is highly regulated, via the TVA reservoir system, whereas the Ohio system is largely unregulated. From an institutional standpoint, TVA is a focal point for Tennessee (and, to some extent, Cumberland) system operations and planning issues. For the Ohio River, no one agency has comparable responsibility, although the U.S. Army Corps of Engineers (USACE) does have system-wide responsibility primarily as a result of its ownership of navigation works.

2) For the Tennessee River system, TVA operations and planning models such as PRYSM define a clear modeling framework and corresponding boundary conditions/forcings which could be provided by GCIP products. Essentially this information includes future reservoir inflows over a wide range of future time scales, ranging from a few days to months and seasons. Also, temperature forecasts would be important to the operation of the energy systems.

3) Opportunities to support water management in the Ohio River appear to include navigation interests on the main stem and a variety of reservoir operations on some of the tributaries. These opportunities need to be explored in more detail. Benefits to navigation of improved forecast information appear to exist for forecast periods up to about two weeks.

D.5.1 Relationship to Ongoing NWS Activities

Present operational hydrologic forecast models in use at the two NWS RFCs in LSA-E and by water management agencies do not include new representations of vegetation that have been developed by the land surface community, do not model the surface energy budget, and generally make limited use of available soils, land use and remote sensing information. On the other hand the land surface models are beginning to include hydrologic components that account for infiltration, surface runoff, and subsurface runoff and water storage. As GCIP begins to focus on the LSA-E, subsurface storage and runoff processes will need to be represented well in the land surface models. This will be required for these models to represent the surface moisture conditions that actually exist in the LSA-E and that are important for surface forcing of the atmosphere in climate models as well as weather prediction models. On the other hand, operational hydrologic prediction models would be significantly improved if they included better and more physically based representations being developed by GCIP for application in atmospheric models and for use in LDAS to provide initial soil moisture and temperature information for NWP models.

NWS is developing an ensemble precipitation forecasting capability. This will use ensemble forecasts from regional and global numerical prediction models, but it will include a range of statistical approaches to processing model output information, for simulating fine scale space-time characteristics of precipitation not represented in model output, and for accounting for short-term forecast uncertainty that may not be included in NWP ensemble products. This also includes development of a precipitation snalysis system to be used at RFCs that will include various statistical tools for combining all of the information from different sources and for producing the final precipitation ensembles for the hydrologic models.

D.5.2 Relevance of GCIP Plans to Water Resources Operations in LSA-E

TVA has an interest in streamflow forecasts with two lead times: a) for operational purposes (up to about a week); and b) for planning purposes (months to seasonal). At present, TVA uses probabilistic (10, 50, 90 percentile) forecasts derived from NCEP products; these are used as forcings in the Lettenmaier/Grygier/Stedinger model streamflows (Sacramento model for five index catchments disaggregated stochastically to 42 inflow nodes). For planning purposes, an analogue approach is used, wherein historical observed streamflows for selected years are routed through a reservoir system model. In addition to inflows to the reservoir system, TVA has an interest in forecasts of surface air temperature, which affect both water temperature, which is a key operating constraint, and power demand.

The PRSYM model was implemented by a research group, and is not currently used operationally by TVA. The ESP approach is not used operationally at present in LSA-E, either by the NWS River Forecast Centers, or by TVA. There is a potential TVA interest in ESP-type forecasts over a range of time scales from several days (for power operations purposes) through seasonal (for power planning).

The NWS scheme(s) for producing QPF are evolving. For short lead times (out to about two days), forecasts will be produced from Eta model output. Because the source of forecast uncertainty is not entirely clear at short lead times (probably a combination of uncertainty in model initialization, parameter error, and residual error due to subgrid effects) it will be necessary to develop schemes to represent, possibly via rescaling, forecast error probability density functions. At longer time scales (up to two weeks), ensemble forecasts will be produced using the NCEP's global model. At these lead times, ensemble predictions are expected to represent more realistically the range of likely forecast errors. Finally, at seasonal time scales, ensemble forecasts will be developed from NCEP's coupled ocean-atmosphere model.

D.5.3 Recommendations

Improvements in short and long-range weather forecasting represent the strongest tie between the GCIP research community and water resources operations, both generally and for LSA-E in particular. As a means to direct the LSA-E water resources activity in this direction, the feasibility of developing an experimental water resources forecast capability for part or all of LSA-E was recommended, as follows:

1) GCIP should develop an experimental streamflow forecast capability for the two major river systems within LSA-E: The Tennessee-Cumberland, and the Ohio River systems. It is important that this activity be implemented with parallel research and operational pathways, the latter of which would incorporate the involvement of the two RFCs that operate in LSA-E. This capability may well encompass multiple modeling systems, but should have the following general attributes:

a) For the Tennessee-Cumberland River systems, produce streamflow at inflow points to existing TVA reservoir systems models, such as the PRSYM system developed collaboratively between TVA, USGS, and other cooperators;

b) For the Ohio River System, forecast points should be selected to match those used by NWS/OHRFC;

c) The system should have the capability of using off-line (e.g., observational) forcings, as well as forecast products produced by the NCEP models.

d) Hydrologic developments should be undertaken as a cooperative effort with the two NWS River Forecast Centers, as well as the key operating agencies (TVA in the case of the Tennessee-Cumberland system; USACE in the case of the Ohio);

2) An ensemble approach to hydrologic forecasting is needed for several reasons. First, PRYSM-type water resources systems models are designed to process ensembles of events to evaluate the implications of alternative operating decisions when the future reservoir inflows are not known exactly. In other words, PRYSM-type models need ensemble forecasts of reservoir inflows. In addition, ensemble prediction methods allow uncertainty in future precipitation patterns throughout a river basin to be analyzed in a way that is statistically consistent for all forecast points in the basin. The TVA system could provide an excellent test site for evaluation of ensemble hydrologic forecasts derived from coupled land-atmosphere models. In this context, analysis of precipitation climatologies should be undertaken to support verification and testing of precipitation forecasts, including ensemble precipitation forecasts. In addition, hydrologically relevant verification methods are needed to assess precipitation forecasts. This includes techniques to assure that the climatology of precipitation forecasts (including ensemble forecasts) matches climatology (i.e. the forecasts are statistically unbiased). Also, hydrologically relevant approaches are needed to measure the skill in these forecasts over a range of space and time scales.

3) Opportunities for diagnosis of NWP models' soil moisture should be exploited using the parallel simulations produced using observed forcings. The potential for updating for NWP model soil moisture using streamflow prediction errors should be evaluated as well.

4) Consideration should be given to broadening the scope of the proposed GCIP/Tennessee River workshop to include some aspects of the Ohio River as well, especially synergisms in the operation of these two systems with respect to effects on the Lower Mississippi River.

5) Attention should be given to the role of biases in both meteorological forecasts (forcings to hydrologic forecast models) and in the hydrologic models themselves. Every hydrologic model includes at least some seasonal bias in the statistical properties (e.g., means and variances) of model outputs when the models are operated in a simulation mode using historical observations. Some method of correcting for these biases is essential for water resource applications of the forecasts. The required corrections usually must be accomplished through post processing of model outputs. Experiments are needed to demonstrate that the climatology of hydrologic forecasts agree with the climatology of historical streamflow events. In addition, useful methods to measure the skill in these forecasts need to be demonstrated to develop the appropriate level of confidence among water resource managers.

D.6 Research Issues Work Session

This Work Session used the results from the first set of Work Sessions to develop an overall listing of the research topics which GCIP should concentrate on during the period of 1997 and 1998 for focused studies on cold season/region hydrometeorology in the LSA-NC. It was agreed that:

1) LSA -E has a wide array of precipitation regimes influenced by orography, soil moisture, and land use.

2) A large question for coupled modeling within the LSA -E is how can models be applied to such things as areal averaging across the region.

3) The LSA-E has high temporal variability in precipitation as well as the highest precipitation within the GCIP region as a whole. Additionally, the LSA-E has systemic wet and dry periods that have a pronounced effect on hydrometeorology.

4) Surface energy balance/radiation data are sparse across the LSA-E, but could be very useful for coupled modeling if the existing sites are augmented.

The following items were recommended:

Augment surface flux capabilities within the LSA-E at specific sites selected for focus studies.
Investigate the availability of aircraft measurements within the LSA-E.
Develop an action plan for evaluating and improving WSR- 88D and gauge precipitation data sets for model prediction (e.g., topography, snow cover)

One other aspect that needs to be undertaken is to evaluate and improve GOES and polar orbiting data for surface radiation budgets, radiative flux estimates, and to develop data sets for flux profiling of surface fluxes. It was suggested there be development of the LDAS concept, both for operational and research uses, and, to develop a strategy to validate with streamflow gauging with emphasis on focus study areas.

It was recommended that GCIP/DACOM include the following sites in their inventory of data available in the LSA-E.

1. Walker Branch Watershed at Oak Ridge

2. Bondville, IL SURFRAD site/Reifsteck farm in situ site

3. USDA-ARS Hydrologic Experiment Station at Coshocton, OH

4. Alabama A&M University research farm and U.S. Army Redstone Arsenal Meteorological station, Huntsville, AL

5. Panola experimental watershed near Atlanta, GA

6. Giles County, TN -- TVA Land Between the Lakes site

7. Coweta Experimental Watershed, Otto, NC

Additionally, land-grant universities within the LSA-E (i.e., agricultural schools) should be contacted to find out if they monitor any flux tower sites and instrumented watersheds within the LSA-E. Potential schools are: University of Tennessee, Knoxville; University of Kentucky; University of Georgia; Auburn University; Mississippi State University; Ohio State University; West Virginia University; Virginia Tech as well as possibly others.

D.7 Data Issues Work Session

This Work Session used the results from the first set of Work Sessions to develop a consolidated list of data requirements for the LSA-E. The Work Session started with the "strawman" list of data requirements which had been developed prior to the workshop. Several possible additions of data from states within and just outside the LSA-E were discussed. This included the Georgia Forestry Commission (28 meteorological stations), the Alabama Weather Observing Network (several automatic meteorological stations) and Alabama Redstone (18 meteorological stations), the North Carolina State Network (14 meteorological stations). Possible additions to upper-air data include profiler data from Redstone Arsenal, University of Alabama-Huntsville (UAH) and Oak Ridge, Tennessee.The consolidated list which resulted from a discussion in a Plenary Session at the Workshop was given in Section 10 of this report.

The group recommended the following actions for GCIP in preparation for research activities in the LSA-E:

Perform a survey to find out what data products are available and what instrumentation is available within the LSA-E. Focus on existing data sources and data sets within the LSA-E.
Produce a detailed survey of in situ data availability within the LSA-E
Identify researcher requirements for WSR-88D data (i.e., volumes, cost, browse availability).

The group raised a number of questions pertaining to the availability and use of satellite remote sensing data in the LSA-E.

What is the future of the satellite data source module as part of the Data Management and Service System? What happens to data availability after the MSFC DAAC closes?

What are the satellite data requirements for GCIP researchers?

What is the quantity of data available? (How accessible are these data and at what cost?)

Is there a need for a satellite data source module and what role should it play in LSA-E research? (e.g., as a provider/pointer?)

The Session was informed that the MSFC/DAAC as the current satellite remote sensing data source module Work is developing a detailed survey of data availability through remote sensing satellites affecting the LSA-E.