6. CRITICAL VARIABLES

A number of variables are critical to the success of GCIP and were designated as Principal Research Areas for GCIP. Each of these are described in this section in terms of research activities needed by GCIP and the plans for data products to support GCIP research activities.

6.1. Precipitation

Note: A Pilot Project involving studies of Orographic Precipitation in the LSA-NW and specifically the Black Hills region of South Dakota is described briefly in Section 7.4.2.

6.1.4. Precipitation Predictability

OBJECTIVE:Assess the limits of predictability of atmospheric model precipitation as a function of scale.

Activities to support this objective are:

• Understand the effects of relative patterns of convective/stratiform rainfall and of subgrid scale spatial rainfall variability on rainfall prediction at the atmospheric model grid scale and temporal scales of hours to days.

• Understand how parameterizations of cloud microphysical processes affect precipitation prediction at the atmospheric model grid scale and temporal scales of hours to days.

• Investigate the accuracy of rainfall predictions at the monthly, seasonal, and interannual scales.

• Understand how the resolution of orography affects precipitation predictions,which affect hydrologic balances and flooding over the Mississippi river basin.

• Develop innovative approaches for validating rainfall predictions from atmospheric and coupled models at a range of space-time scales of special interest and quantify the degree to which models capture important physical and statistical features/signatures that there is evidence for, from observations. This sort of validation can provide guidance for improved cloud process parameterization in coupled modeling.

• Determine the inherent limits of predictability of precipitation, by studying the sensitivity of predictions to initial and boundary conditions, and in particular in a nested modeling environment, and also study the scales at which it is ``better'' to nest.

• Develop methods for validating ensemble predictions.

• Develop methods for integrating information at different scales i.e., point observations versus remote sensing measurements and model outputs which are space-time averages over varying scales.

The contact person for this archived precipitation analysis data product is Roy Jenne at NCAR (e-mail: Jenne@ucar.edu).

The objective of the precipitation observation composite is to provide a quality controlled composite of all available precipitation gauge observations in a common format. The data product contains precipitation data from all real-time and recording gauges in the geographic domain as both hourly and daily totals. The Composite is produced by the In-Situ Data Source Module using data from up to 14 different observing networks. A precipitation observation composite was produced for each of the GCIP Initial Data Sets. Evolutionary improvements in quality control procedures will be implemented as proven techniques warrant. There are no current plans to correct for measurement errors by the different sensor systems. However, it is expected that any adjustments for measurement errors could be done using this precipitation observation composite data set. The contact person for this archived precipitation observation composite data product is Steve Williams at UCAR/JOSS (e-mail: sfw@ncar.ucar.edu).

6.1.8 Snow Measurements and Analysis

Point snow measurement relies primarily on the Natural Resources Conservation Survey SNOpack TELemetry (SNOTEL) network, which is largely to the west of the Mississippi River basin, and a comparatively sparse network of snow depth measurements at NWS synoptic stations. Snow courses are measured by various agencies, but these are limited and are restricted to the higher snowfall areas.

Remote sensing offers a more practical approach to assess snow over large areas. However, the need for new techniques or additional ground truth measurements has to be considered. The program in NESDIS is focused on the development of an interactive system for producing daily, rather than the current weekly, Northern Hemisphere snow maps on Hewlett Packard 755 UNIX-based workstations from a variety of satellite imagery and derived mapped products in one hour or less. Resolution of the final product will be improved from 190 Km to 23 Km. The final product will also provide information on snow depth in addition to snow cover.

GCIP is planning to derive adjusted values for in-situ solid precipitation measurements compiled for ESOP-97 and ESOP-98 based on the results of studies by E. Peck and P. Groisman now underway.

6.2 SOIL MOISTURE

OVERALL OBJECTIVE: Improve understanding and estimation of the space-time structure of soil moisture, the relationship between model estimates of soil moisture and observations of soil moisture, and to produce soil moisture fields for the GCIP area to be used as diagnostic and input data for modeling.

6.2.1 In Situ Soil Moisture Measurements

A survey by the Natural Resources Conservation Service at the time GCIP was preparing its implementation Plan in 1992-93 revealed that there were very few soil measurement sites in the Mississippi river basin. A network operated by the Illinois State Water Survey could provide measurements on a weekly schedule during the crop growing season and biweekly during the remainder of the year.

GCIP started an effort in 1994 to enhance the soil moisture measurements both in number of sites and frequency of measurements by providing some support to the ARS experimental site in Little Washita Watershed to analyze a set of automated soil moisture profile measurement systems and to install some test sites in the watershed. This small evaluation task has evolved to a rather extensive network of soil moisture and soil temperature profile measurement sites in the LSA-SW.

Six soil moisture sensing systems were installed in the Little Washita Watershed in the summer of 1995. An additional seven sensor systems were installed in this Watershed during 1996. A total of 22 soil moisture sensing systems were installed within the ARM/CART site. The first seven were installed and operating by the beginning of ESOP-96 in April 1996 and the remaining were installed by April 1997. An example of the relative soil moisture response curves in the ARM/CART site is given in Figure 6-1 which was very dry during the spring and early summer. The Campbell Scientific Heat Dissipation Soil Moisture Sensor (Model 229L) provides data from six different depths as shown in Figure 6-1. The calibration to convert the sensor is not yet completed. Therefore, the relative response in degrees celsius is given in the figure with lower values wetter and higher values drier. The curves from Ashton in May 1996 are typical of the response from many sites this spring and summer. The soil was very dry throughout the profile, and what little rain fell did not infiltrate very deeply into the profile. At Ashton, the rain on May 10th wetted the top two sensors, with only a slight amount of moisture penetrating as far as the 35-cm sensor. The Oklahoma Mesonet installed soil moisture sensing systems at about half of their 109 stations in the state-wide mesonetwork. There are plans to extend the soil moisture measuring systems to all of the 109 sites in the network. The situation in the LSA-SW is such that GCIP can potentially compile in-situ soil moisture measurements on three different scales using automated soil moisture sensing systems

Figure 6-1 Relative soil moisture response curves for Ashton, OK during May 1996 from the Campbell Scientific Heat Dissipation Soil Moisture Sensor.

The Southern Great Plains Experiment conducted in June-July 1997 was an intensive observing period focused on measuring and mapping soil moisture. Further details on this experiment are provided in Section 6.2.5

Additional in situ soil moisture measurements throughout the GCIP region should be encouraged, especially in the LSA-E and LSA-NW. The in situ measurements are necessary to document the seasonal and interannual variability in addition to providing index measurement sites for the validation and continued evaluation of model estimates of soil moisture discussed in the following section.

6.2.2 Soil Moisture Fields

OBJECTIVE: Produce the best possible estimates of soil moisture at four depths over the entire GCIP study area with the initial emphasis over the LSA-SW.

Activities that are needed to support this objective are:

• A validated soil moisture product is needed for the Mississippi River Basin at a spatial scale of about 40 km and a daily temporal scale for four depths corresponding to the Eta and MAPS model output. This assimilated product must be produced from a variety of data sources, including output from hydrologic models driven by measured meteorological data, in situ soil moisture observations, the NOAA/NWS gamma flights and satellite remote sensing. The currently available remote sensing data are from the SMMR and SSM/I satellites and, although not ideal for a soil moisture measurement, do have a soil moisture response over much of the GCIP area. The challenge will be to combine the various data forms to produce the "best" possible gridded product and to develop a way to validate the product with in situ data, preferably data not used in the assimilation process.

• Initially, a subset of the soil moisture product should be developed for the LSA-SW because this is the area where the most in situ data are available and the region where current remote sensing can provide the best information because of the relatively less dense vegetation cover.

• A second subset of the soil moisture product needs to be developed for the LSA-NC. Here the Illinois in situ soil moisture data set can be used for validation and/or assimilating the data set. The issue of cold season hydrology and frozen soils needs to be addressed with this data set.The development of methods to combine remotely sensed and in situ soil moisture should be encouraged. Of particular interest are methods that are accurate at the beginning of the cold season, just before the soil freezes and snow cover commences, and just after snow cover has disappeared. An example of this is the combining of aircraft and in-situ measurements made at the beginning of the cold season by the NOAA/NWS - National Operational Hydrology Remote Sensing Center . In addition, passive microwave data should be useful for delineating frozen and non-frozen soils.

Surface cover in the test area during the experiment time frame from June 18 to July 16, 1997 was predominantly senesced or harvested winter wheat and rangeland pasture. Several significant soil moisture dry downs occurred at different times in different parts of the test area due to thunderstorm activity. Additional information about SGP97 can be found on the World Wide Web at the URL address: http://hydrolab.arsusda.gov/sgp97/

ESTAR brightness temperature maps of the SGP97 area are currently undergoing detailed reprocessing and registration. It is anticipated that soil moisture maps derived from the ESTAR data will made available to the scientific community near the end of 1998.

Future Satellite Sensors

Within the near future, there will be several space borne instruments that will contribute to the technology of remote sensing of soil moisture. The japanese are currently building two identical passive microwave instruments, AMSR (Advanced Microwave Scanning Radiometer) that will have a C-band radiometer (6.9 GHz) in addition to other microwave bands that match the SSM/I bands. The first instrument will be launched on the Japanese ADEOS-II (Advanced Earth Observing Satellite-II) in 1999 and the second will be launched on the NASA EOS PM platform in 2000. The C-band instrument will provide useful data for soil moisture at a spatial resolution of about 50km and a three to four day repeat cycle.

The first opportunity for an L-Band instrument may come in the 2002 to 2003 time frame through an ESSP (Earth System Science Pathfinder) program. The challenge for an L-Band instrument is to erect a very large antenna in space in order to achieve useful spatial resolution. Several concepts are being studied and one or more will be proposed in 1998. The ESTAR instrument flown in the SGP-97 is an airborne prototype of a likely space borne instrument and the previous flights with this instrument have provided a large amount of data and experience in imaging processing and algorithm development.

6.2.6 Recommended High Priority Activities

The Soil Moisture Research Area has very little ongoing research or activities that can be specifically attributable to GCIP, with a major exception being the support for installation of in situ soil moisture stations. The needs for soil moisture have been expressed by a number of other GCIP Research Areas, e.g. Coupled Modeling Research in Section 2, as well as individual researchers. The following items were identified in the GCIP Soil Moisture meeting (in Boulder in November 1997) as being essential to the successful implementation of the entire GCIP program:

• Variability of soil moisture; understanding the relationship between point measurements and areal representations AND the relationship between surface measurements and the total profile soil moisture.

• The need to develop validated, daily soil moisture fields for the entire GCIP area at a 40 km grid and four depths.

• To develop the cold season products of soil moisture and temperature fields, including frozen and unfrozen soils.

Three specific steps were recommended to accomplish this:

• Coordination among the various funding agency program managers, both in the inclusion of these three specific needs in future research announcements and in the budgeting of sufficient funds to do the job.

• Recognition that we can measure soil moisture with existing satellite borne sensors.

• Use the SGP-97 data sets and activities as the spring board to accomplish these goals.

Long term activities that should be started within the next two or three years includes:

• Increase the number of in situ monitoring stations to include additional soil/climate regimes. The North - South transect along 96W longitude represents a good start.

• Start using existing remote sensing capabilities and anticipate the infusion of AMSR and hopefully L-Band data.

• Interact with the coupled modeling community.

OVERALL OBJECTIVE: Improve the quantitative understanding of the relationships between model parameterizations of land processes and land surface characteristics; and facilitate the development, test, evaluation, and validation of multiresolution land surface characteristics data and information required by GCIP researchers for developing, parameterizing, initializing, and validating atmospheric and hydrological models.

6.3.1 Land Surface Characteristics Research

The strategy for this land surface characterization research is twofold. In the near term, the primary emphasis is on facilitating the adaptation, tailoring, test and evaluation, and validation of existing land surface characteristics data sets that will meet the immediate requirements of GCIP's Principal Research Areas. The multiresolution land surface data requirements of GCIP researchers will be documented and the GCIP land surface characterization research plan will be updated based on regular feedback from GCIP modelers, as well as research results concerning land process modeling activities of PILPS and ISLSCP. This near-term strategy also includes adapting and testing promising biophysical remote sensing algorithms that are available in the literature, for example published results from ISLSCP's remote sensing science activities involving FIFE, Boreal Ecosystem Atmosphere Study (BOREAS), or the GEWEX/ISLSCP global one-degree latitude-longitude global land data sets published on compact disk, read-only memory (CD-ROM). Many GCIP modelers will conduct land characterization research as an integral part of their efforts to develop land surface process models and parameterizations, therefore, facilitating the cross-disciplinary flow and sharing of land characterization results and information within the GCIP research community is needed. GCIP's longer-term strategy for land surface characterization research will focus on developing and testing enhanced high-resolution land data sets. This includes collecting field data that are necessary to develop, adapt, test, and validate promising remote sensing algorithms for land cover characterization and model parameterizations; conducting advanced remote sensing research, for example, canopy reflectance modeling; and investigating landscape heterogeneity, grid cell aggregation rules, and land data interrelationships as related to land process parameterizations. This longer-term strategy also includes provisions by GCIP to test and evaluate remote sensing data sets that will become potentially available at as yet unknown dates following the planned launches of the NASA-led Earth Observing System (EOS) AM1 Platform and Landsat 7 during the mid-1998 time-frame.

Multiresolution land surface characterization research in the near-term will be directed towards meeting the minimum requirements of GCIP Principal Research Areas for land cover, soils, and topographic data, including associated characteristics and properties of each, at four regional scales. For example, the initial project regions and their associated gridding intervals included the CSA and LSA-SW (30-km grids), ARM/CART as the initial ISA (10-km grid), and Little Washita as the initial SSA (4-km grid). The primary land surface data sets that were available throughout the conterminous United States to meet some of GCIP's early requirements for land data within these four regions included various 1-km and coarser spatial resolution, advanced very-high-resolution radiometer (AVHRR) data products from NOAA's polar-orbiting satellites; the 1:250,000-scale USDA/Natural Resources Conservation Service State Soil Geographic Database (STATSGO); and DEMs of 0.5-km and approximately 100-m grid cell resolutions, respectively, available from the USGS. Land characterization research focused on the adaptation and use of these primary data sets as the basis to develop, test, and evaluate key derivative land surface characteristics data sets for use by GCIP modelers.

As GCIP evolves, land surface characterization research will focus on meeting the changing requirements of coupled modeling in GCIP and the testing of land surface data in newly defined LSAs. For example, land surface parameterization sensitivity studies by PILPS and GCIP investigators have helped to identify critical requirements for detailed soils information and fractional vegetation cover data (percent bare soil .vs. percent vegetation) as key inputs to land surface parameterizations. The GCIP research activities for the LSA-NC began in 1997 with research planning for the LSA-E (1998-1999) in the final stages. The GCIP research planning for the LSA-NW (1999-2000) is scheduled to begin in early 1998. The GCIP research at the CSA and LSA scales will benefit from land cover characteristics data derived from remote sensing algorithms developed as part of ISLSCP Initiatives (No. 1 and No. 2) or, as part of EOS AM1 project activities, when available.

Higher resolution land data sources are need for ISA-scale and small watershed regions in GCIP. Examples of these subregions include the ISA ARM/CART, the Upper Walnut River watershed located within the ARM/CART as part of the Cooperative Atmosphere-Surface Exchange Study (CASES) project, and the SSA Little Washita watershed located just to the south of the ARM/CART. Candidate SSAs within the LSA EAST include the river subbasins within the Tennessee River drainage basin and the Goodwin Creek watershed (part of the Yazoo River basin), a USDA/ARS experimental watershed located in north central Mississippi. Similar ISAs and SSAs for the LSA-NC and the LSA-NW are yet to be determined.

Some of the key secondary land data sources could include various types of 30-m LANDSAT thematic mapper (TM) data products for land cover characterization within the ISA- and SSA-scale regions, selected county-level digital USDA/Natural Resources Conservation Service Soil Survey Geographic Database (SSURGO) (as available), USGS digital 60-m DEMs for the ARM/CART, and USGS 30-m DEMs available in a 7.5-minute quad format for selected locations within the GCIP domain. The land data sets developed for the Upper Mississippi region by the Scientific Assessment and Strategy Team (SAST) concerning flood plain management following the 1993 floods potentially represent a significant contribution to the land surface characterization requirements for the LSA-NC and LSA-NW (see the World Wide Web at the URL address: http://edcwww.cr.usgs.gov/sast-home.html). Detailed analysis of multiresolution satellite data for the ISAs, for example the ARM/CART region, can contribute improved remote sensing algorithms that can be applied within the LSA- and CSA-scale regions.

Additionally, the identification and facilitation of the use of appropriate data analysis tools, such as GISs and digital image processing systems, will be needed to tailor land surface characteristics from primary data sets and to integrate and analyze disparate data sets of interest to land process researchers. Both standard and new image processing techniques will be necessary for analysis of multitemporal land cover characteristics data, frequently available from satellite remote sensing systems with different spatial resolutions. Moreover, the application of appropriate geostatistical techniques, such as measures of dispersion or aggregation of landscape patterns, will be investigated to assist in understanding the spatial linkages extant between land surface characteristics and the hydrometeorological conditions within the GCIP study area.

Land surface characterization research is highly interdisciplinary in scope. Therefore, an equally important high-priority task is to develop Federal agency participation and resource support for cooperative work on the accomplishment of GCIP's land surface characteristics research objectives and activities. Some of the potential Federal agency participants for conducting and supporting this land surface characterization research include NOAA (NWS and NESDIS), the USGS (National Mapping Division and Water Resources Division), NASA [Marshall Space Flight Center (MSFC) and GSFC] and the USDA [ARS, Natural Resources Conservation Service, and National Agricultural Statistics Service (NASS)]. In many cases, the results of this interdisciplinary land surface characterization research will directly benefit agency missions, such as those concerning land data set development, remote sensing science, operational programs involving atmospheric and hydrological modeling, natural resource assessment, and agricultural monitoring and forecasting. Furthermore, activities such as SAST, involving flood disaster management, can contribute to GCIP both in terms of a supplier of land data and as a key user of GCIP atmospheric, hydrologic, and water resource products for policy decision making. The efforts of such Federal agencies would complement contributions made by GCIP's research community including expertise at universities. The coordination of this research with potential contributions by GEWEX/ISLSCP presents an outstanding opportunity, especially for biophysical remote sensing algorithm development, operational data set development, and scaling research.

This land surface characterization research strategy will be accomplished through objectives and associated research activities involving land cover characteristics, soils and geology, and topographic information. The research activities under each objective are listed according to priority for accomplishment.

6.3.2 Land Cover Characteristics and Associated Data Products

The biophysical remote sensing and land-atmosphere interactions modeling communities are currently addressing many of the research questions and related data development issues concerning the potential role of land cover characteristics as determinants of land surface processes. This research by atmospheric and hydrological modelers is concerned with understanding and parameterizing the effects of land cover characteristics in their models and parameterizations (i.e., land cover and vegetation type, land use, the physical and biophysical properties of vegetation including the temporal dynamics, and more recently the spatial heterogeneity of the landscape). In many cases, these two communities also share common interests in developing the experimental remote sensing algorithms that are needed to estimate or derive various types of land cover characteristics from satellite data over large areas. Examples range from the use of multitemporal satellite-derived spectral vegetation greenness indexes for land cover classification and estimating leaf area index (LAI) to more advanced canopy reflectance modeling for estimating biophysical parameters and processes. Facilitating the adaptation and use of published research results and biophysical remote sensing algorithms within GCIP is a key requirement.

Some of the sources for land cover characteristics data include the global land data sets for land-atmosphere interactions modeling published on CD-ROM by NASA/GSFC under GEWEX/ISLSCP Initiative No. 1, plus various AVHRR data sets produced by NASA, NOAA/NESDIS, and USGS. For example, NASA's ISLSCP Initiative No. 1 CD-ROM includes monthly one-degree by one-degree latitude-longitude calibrated, continental-scale NDVI data (1987-88); enhanced NDVI fields; Fraction of Absorbed Photosynthetically Active Radiation (FPAR) fields derived from enhanced-NDVI data; LAI and canopy greenness resistance fraction calculated from the derived FPAR fields; surface albedo and roughness length fields derived from land process models; and canopy photosynthesis and canopy conductance fields estimated by inverting the Simple Biosphere Model (SiB2) land surface parameterization (LSP) with FPAR as the key model input. A key step in the biophysical parameter estimation was the development of the "Fourier-adjusted, solar zenith angle-corrected, interpolated and reconstructed" (FASIR) algorithm to derive the enhanced-NDVIs. The CD-ROM also includes a one-degree global land cover data set developed by the University of Maryland. Overall, this ISLSCP CD-ROM contains the first set of global land cover and land cover biophysical parameter data that are derived in an internally consistent fashion.

Although these ISLSCP Initiative No. 1 CD-ROM data are of direct interest to GCM and coarse grid cell resolution mesoscale modeling, the remote sensing algorithms and approaches for processing satellite reflectance data and inverting an LSP to derive the land cover characteristics can guide similar data set development efforts using higher resolution AVHRR and LANDSAT TM data. The FASIR algorithm can be adapted for developing LSA-scale data sets for test and evaluation in GCIP. NASA/GSFC is currently leading the development of new global consistently-derived data sets under the ISLSCP Initiative No. 2 activity which is focusing on enhanced global land cover characteristics data sets at a 1/2-degree latitude-longitude grid for the ten year period, 1986-1995. The ISLSCP No. 2 data are planned for release during the 1998-99 timeframe. One source for this multi-year global analysis is the 8-km AVHRR Global Area Coverage (GAC) Pathfinder data set developed jointly by NASA and NOAA for the period 1982-1995. This global 8-km data set and the ISLSCP Initiative No. 1 global data can be obtained via the NASA/GSFC DAAC WWW site ( http://daac.gsfc.nasa.gov).

The NOAA/NESDIS has developed several AVHRR global vegetation index (GVI) data sets. These data sets include weekly satellite image composites consisting of five AVHRR channels, solar zenith and azimuth angles, and the GVI for 1985 to the present. These data are calibrated for sensor drift and intersensor variability, and are available in a 1/6-degree resolution latitude-longitude global product. NOAA/NESDIS has produced a five-year climatology of average GVI data for the globe. More recently, NOAA/NESDIS has developed a NDVI-scaled "fraction of green vegetation index" ( http://orbit-net.nesdis.noaa.gov:80/ora/lst/gutmanpage.html). This data set is currently undergoing test and evaluation in the NOAA/NCEP Eta model. NOAA/NESDIS has also investigated the use of GVI data in vegetation crop indexes as a tool to detect and monitor large-area meteorological drought. Finally, the NOAA/NESDIS National Geophysical Data Center recently released Disk B of the Global Ecosystems Database that includes the Fedorova et al., World Vegetation Cover and the Bazilevich Global Primary Productivity.

The USGS EROS Data Center (EDC) has developed 1-km AVHRR data sets for the conterminous United States and is now processing global 1-km AVHRR data for land areas. The data sets for the conterminous United States include biweekly AVHRR time-series image composites on CD-ROM (1990-1996) and a prototype land cover characteristics data set for 1990 on CD-ROM. Ongoing USGS activities for the conterminous United States include the development of experimental, temporally smoothed 1-km seasonal NDVI greenness statistics for test and evaluation. These statistics consist of 12 seasonal characteristics that are associated with each 1-km NDVI seasonal profile for each year during the period 1989 to 1993, as well as the five-year means throughout the conterminous United States.

Under the auspices of the International Geosphere-Biosphere Project (IGBP)-led 1-km AVHRR global land cover data set development activity, the USGS is currently processing global, 10-day AVHRR image composites for land areas. Prototype 1-km AVHRR land cover data sets for the North American continent were developed as part of a global land cover mapping effort. These land cover data for North America include individual data sets for the BATS, Sib2, IGBP, and other land cover classification schemes plus associated monthly AVHRR image composites and a 1-km digital elevation model (DEM) for North America. These data sets can be accessed online via the EDC Distributed Active Archive (DAAC) Home page (http://edcwww.cr.usgs.gov/landdaac/). The 1-km AVHRR IGBP global land cover data are currently undergoing validation as part of an independently-led IGBP project activity. Several global climate change research modelers, including some GCIP investigators, are currently testing and evaluating these USGS data sets.

In mid-1998, the Earth Observing System (EOS) AM1 platform is scheduled for launch as part of NASA's Mission to Planet Earth (MTPE). A wide variety of land cover characteristics data are scheduled to be produced from data collected by the MODIS, MISR, ASTER, and CERES sensors on board the AM1 Platform. When ready for test and evaluation at some later date, these new data sets would be an important contribution to GCIP research investigations. For example, enhanced atmospherically-corrected reflectance data and spectral vegetation index data would be potentially available. In addition, current NASA plans also call for the 1998-launch of Landsat 7, which will be in near-synchronous orbit with the AM1 package. Land surface research will benefit from concurrent overlapping Landsat 7 and EOS AM1 products. Further information is available from the NASA Mission to Planet Earth WWW page (http://www.hq.nasa.gov/office/mtpe).

OBJECTIVE: Improve the quantitative understanding of the relationships between land cover characteristics and the land surface parameterizations and land process components of atmospheric and hydrological models, and meet the requirements of the GCIP modeling and research activities for multiresolution land cover characteristics data.

Activities in support of this objective in order of priority will:

• Define the requirements of GCIP modelers for multiresolution land cover characteristics data, document available data sources, and assess the adequacy of available data for GCIP Principal Research Areas based on interviews with GCIP PIs. Ongoing feedback is needed from GCIP, PILPS, and ISLSCP activities concerning requirements for land cover characteristics data and the results of model sensitivity tests.

• Facilitate the test and evaluation of recently developed land cover characteristics data sets at the continental scale including new USGS 1-km AVHRR land cover and NOAA/NESDIS 1/6-degree latitude-longitude fractional vegetation greenness index data.

• Develop multiresolution, aggregated land cover characteristics data sets that are internally consistent and collectively "harmonized" within model spatial domain grid cells. Products will include, at 10 km and 30 km grid cell sizes, the three predominant land cover/land use classes in each grid cell. Products will also include a climatology of the fractional presence of green vegetation associated with the three predominant land use/land cover classes. Products for the conterminous United States will be based on a five-year climatology, while the prototype global product will be derived from an 18-month time series.

• In addition to data quality and evaluation criteria, this research also requires investigation of various aggregation rules, such as aggregation by predominant land cover class or aggregation that retains the percent composition in each grid cell. For example, the Eta model's land surface parameterization now has a requirement for 10-20 km grid cell aggregations of land cover (dominant and secondary classes), multi-year average NDVI, fractional vegetation greenness index, albedo, and other biophysical parameters that are internally consistent with each other from a land surface process perspective within each Eta model grid cell. Mesoscale atmospheric and macroscale hydrologic basin-scale models require the incorporation of soils properties and topographic features. These types of land surface data sets are needed at multiple resolutions to meet the nested grid requirements in this GCIP coupled modeling approach.

• Deliver enhanced land surface characteristics data sets and place them on-line for easy access. Improved seasonal biophysical parameters at 10-20 km grid cell aggregations are needed by coupled modelers in GCIP. These include improved spectral vegetation indexes, land surface albedo, surface roughness, fractional vegetation coverage (percent vegetation .vs. percent bare ground), LAI, FPAR, and "wet" biomass for passive microwave soil moisture research. Requirements range from basin-scale modeling to GCIP research in the LSAs, especially the LSA-East and LSA-NW. In general, physically-based algorithms are an essential requirement for biophysical parameter estimation at least on monthly time scales. For example, the NASA-led ISLSCP/GEWEX Initiative Number 2 could provide enhanced biophysical land cover land cover parameters at a 0.5 degree grid size based on the FASIR algorithm. Adapting this FASIR algorithm approach for application at a 10-20 km analysis is another option. These GCIP research efforts would be conducted in parallel with and complement the development of land cover characteristics algorithms by NASA-science teams associated the EOS AM1 spacecraft.

• Conduct land cover characterization research with 30 m land cover data sets developed from Landsat TM imagery as part of a USGS- and EPA-led multiresolution land cover characterization activity. Preliminary 30 m land cover data are scheduled for completion in the LSA-East by late 1997 with completion for the remainder of the conterminous United States planned by the end of 1999. These high-resolution land cover data could benefit GCIP LSA-scale research activities involving multiresolution aggregation, scaling, and validation studies. In addition, the DOE ARM program has identified available high-resolution land cover and land use data sets for the ARM/CART, while the USDA/ARS is presently completing a GIS for the Little Washita including recent Landsat TM-derived land cover classifications, historical Landsat MSS products back to 1972, and other land surface characteristics data sets.

• Use GCIP findings to assess the new levels of understanding concerning the role of land surface characteristics in land surface parameterization research.

• Facilitate the use of GCIP model output and data assimilation products by remote sensing data centers to improve remote sensing processing techniques, especially approaches for making atmospheric corrections to satellite reflectance data for atmospheric water vapor content and aerosol concentrations.

• Conduct GIS-based land surface characterization research studies within the Mississippi River Basin to determine multiresolution interrelationships among land cover, biophysical parameter, soils, and topographic data sets including derived parameter fields. In addition to verification of physically, consistent associations, these studies should focus on the role of landscape heterogeneity in parameterizing land surface processes. Several 1-km data sets now available for the conterminous United States establish the foundation for this GIS-based multiresolution analysis. These include USGS 1-km satellite-derived data sets (AVHRR image composites, extensive land cover data, and 5-year average NDVI statistics), a USGS 1-km DEM, and the PSU/ESSC 1-km CONUS-SOILS data set. Other data such as NOAA/NESDIS GVI-based multi-year average NDVI and fractional green vegetation index can be resampled to 1-km (no increased resolution). Ancillary data sets include the GCIP GREDS CD-ROM and the USGS Companion Data sets for the Conterminous United States on CD-ROM. Several GCIP PI's have expressed interest in a GCIP GIS. These data sets provide an opportunity for climatologists, physical geographers, soil scientists, and others to understand interrelationships in the data.

  One prime reason for this research is to ensure that the land surface characteristics data sets on land cover, vegetation attributes, soil properties, and topography are appropriately and consistently tailored within model grid cells or watershed polygons for model applications. In addition to model sensitivity studies concerning accuracy issues for individual data layers, error propagation analysis will also be conducted to assess the net impact of effectively "overlaying" land cover, soil, and topographic data sets in the model, where these data are characterized by differing levels of accuracies, precision, uncertainties, and other data limitations.

• Research is needed to investigate how the spatial heterogeneity of vegetation (i.e., landscape patchiness) affects model parameterization, especially as related to spatial aggregation of data within model grid cells and polygons, or scaling parameterizations. This land surface characterization research emphasizes the spatial component within the landscape, for example, concerning the arrangement, pattern, distribution, and composition of various land cover types within a region that influence or potentially affect land- atmosphere interactions and hydrometeorological relationships.

• Conduct advanced research on satellite data processing and physically-based remote sensing algorithms. For example, atmospherically corrected satellite reflectance data are needed to help overcome the adverse and variable affects of atmospheric water vapor and aerosols on surface reflectance retrievals. Lack of such atmospheric corrections, as well as bidirectional reflectance distribution function (BRDF) corrections, significantly degrade the use of existing satellite reflectance data to calculate vegetation greenness indexes that can be reliably used to study intra-and inter-annual variability of vegetation greenness,i.e., change detection. Advanced land surface temperature/emissivity and biophysical remote sensing algorithms are need to study land surface energy and surface flux conditions.

• Develop plans to test and evaluate remote sensing data sets that will become potentially available at some date following the launches of the NASA-led Earth Observing System (EOS) AM1 Platform and Landsat 7 during the mid-1998 time-frames. The advanced remote sensing science algorithms under development by the MODIS Land Science Team are of particular relevance to GCIP. In addition, current NASA plans call for near-synchronous orbits of the EOS AM1 and Landsat 7 satellite systems, thereby creating a substantial potential for the complementary operation of coarse- and high-resolution satellite data of interest to some GCIP researchers. By testing and evaluation of MODIS land products and by making comparisons to field data, GCIP investigators could make important contributions to the validation of MODIS products.

• Investigate the relationships of landscape thermal infrared conditions, as derived from satellite data (for example, MODIS, ASTER, and Landsat 7 data), with land cover type, biophysical land cover parameters, the land surface energy budget, sensible and latent heat fluxes, and other land surface processes.

• Contribute to the Mississippi River Basin Experiment (MRBEX) in the Year 2000. Early GCIP efforts for MRBEX could potentially begin in Year 1999 in conjunction with the GCIP LSA-NW which will have the theme of "Land Cover Over the Annual Cycle" associated with sparse land cover and snow. The data collection for MRBEX would begin in 1999 with the actual observation period for Year 2000. The GCIP LSA-NW would be used as a means to scale up to the whole Mississippi River Basin in MRBEX. Selected MODIS land surface data products potentially available could make important contributions in the LSA-NW. Use Landsat 7 products in combination with field observations to assist in the validation of existing land cover characteristics data sets and to make recommendations for revisions.

GCIP-funded research has resulted in the development of the first 1-km multi-layer soil characteristics data set for the conterminous United States (CONUS-SOIL). This data set is based on the STATSGO data and provides soil physical and hydraulic properties (soil texture, rock fragment class and volume, depth-to-bedrock, bulk density, porosity, sand, silt, and clay fractions, available water capacity, and hydrologic soil group) for the 48 conterminous United States. A key element of the functional design requirements behind CONUS-SOIL was to provide the data in map projections and formats that would permit users to more easily integrate soil information into their particular modeling applications.The complete CONUS-SOIL data set was released in February, 1997 (WWW access: http://www.essc.psu.edu/soil_info/). The response to this data set from the environmental modeling community has been extremely supportive and positive. System logs indicate downloads of various portions of the data set at a rate of about one dozen per week. Other forms of feedback, including requests and comments, indicate a measurable level of success for this approach to delivering soils information.

CONUS-SOIL provides the most useful data set for regional-scale analysis; however, GCIP researchers will still require, on a selective basis, SSURGO data for detailed watershed studies and intense field observation programs. Although this data set will not be complete for the entire United States or even the GCIP study area for many years, selected watersheds within the Mississippi basin should have this, or similar coverage, within the EOP. The SSURGO and S%' *sets are linked through their mutual connection to the NCSS Soil Interpretation Record (Soil-5) and Map Unit Use File (Soil-6).

A geologic map of surficial geology for the upper Mississippi River Basin was developed by Dr. David Soller of the U.S. Geological Survey in Reston, VA.

OBJECTIVE: Develop methods for using soil physical property data for GCIP atmospheric and hydrological modeling.

Activities in support of this objective in order of priority will:

• Continue to refine the requirements of GCIP modelers and scientific investigators for multiresolution soil physical and derived hydraulic properties data.

• Further research is needed to more fully incorporate the PSU/ESSC 1-km CONUS-SOIL data (available on the WWW site http://www.essc.psu.edu/soil_info/) into the land surface parameterizations of atmospheric and hydrologic models.

• In addition, ongoing feedback is needed from GCIP, PILPS, and ISLSCP activities concerning data requirements for soil properties and the results of model sensitivity analysis to these properties.

• These requirements will need to come from the GCIP modeling community. Conceivably a broad range of models ranging from detailed, distributed parameter, physically based models to lumped parameter and stochastic models will be used in GCIP activities. Each may require a unique level of detail of soils information. The modeling and data set development (soil science) communities must consult on the nature of these needs.

• Continue to refine CONUS-SOIL and extend the data set to a complete North American product by the addition of soils data for Canada and Mexico. Many of the mesoscale models being used in GCIP require land surface characteristics information for model domains which cover a large portion of North America. Recent developments in the Mexican and Canadian soil surveys have now made it possible to begin the task of developing a "CONUS-SOIL-like" product for the full North American continent. In fact, the need for such a data set has been expressed on a number of occasions by GCIP modelers.

• Facilitate development of SSURGO data sets for selected watersheds within the GCIP domain. This information will be vital for support of intense field observations and campaigns during the EOP.

• Test and evaluate the CONUS-SOIL and SSURGO data in GCIP modeling activities.

• Improve quantitative understanding of SOIL data limitations for developing gridded soil physical and hydraulic properties. Specifically, GCIP researchers need quantitative estimates of the uncertainties inherent in the aggregation and disaggregation of soil properties based on sparse soil field measurements and of the limitations of traditional methods for estimating soil hydraulic characteristics (e.g., hydraulic conductivity/matrix potential) from soil physical properties.

• This activity also entails research to determine the acceptable minimum resolution for gridding SSURGO and CONUS-SOIL data according to soil property and location within the GCIP domain. Research is required to investigate various approaches for generating soils information for models. Sensitivity analyses must be conducted.

• Explore need for and availability of geologic data sets on local and regional scales for use in defining the impact of ground water on land surface- atmosphere interactions.

• The impact of ground water on land surface-atmosphere interactions must be further explored. Typically, the upper 2 to 4 m or less of soil profile has been the focus of concern for the parameterizations of these processes. Locally, however, the link to ground water may be significant. GCIP should support further research on this topic by studies of selected data as geologic properties, structure, and knowledge of their relationship to ground water characteristics are known.

Topographic information includes surface elevation data and various derived characteristics such as aspect, slope, stream networks, and drainage basin boundaries. In general, the requirements of atmospheric modelers for topographic data (i.e., spatial and vertical resolution and accuracies) are much less demanding than the requirements for hydrological modeling. For example, available DEMs for the conterminous United States (0.5 km and approximately 100-m resolution) are generally adequate for most atmospheric modeling. A 60-m DEM derived by USGS from 2-arc second elevation contours is available for the entire ARM/CART region and other selected quads. In addition the USGS EROS Data Center has recently completed the development of a global 1-km digital elevation model (DEM), now available on the WWW (http://edcwww.cr.usgs.gov/landdaac/).

The 100-m DEM is generally appropriate for hydrological modeling in large basins (e.g., greater than 1,000 km2 in area). However, topographic data for small basins down to watersheds are needed at two general hydrological scales: hillslope and stream network. The hillslope scale is the scale at which water moves laterally to the stream network. Available USGS 60 m DEMs derived from 2-arc second contour data are generally available for the ARM/CART region.

Hillslope flow distances vary and may be as great as 500 m to 1 km. Definition of hillslope flow paths and the statistics of hillslope characteristics require surface elevation data at about 30 m spatial resolution. Such data have been digitized by the USGS from 1:24,000 scale map sheets for part, but not all of the Mississippi River basin. Also, stream locations (but not drainage boundaries) are available in vector form for these map sheets. Because 30-m resolution data are not available globally nor in some parts of the Mississippi basin, research is needed to see how well hillslope statistics, that are important to some hydrological models, can be estimated from topographic properties of lower resolution terrain data. Research is also needed to determine how important hillslope information is to hydrological response of the land surface. Because 1:24,000 scale maps are not available globally, research is needed on how best to use remote sensing techniques as part of a sampling strategy to develop regionalized hillslope statistics (which may be mapped at an appropriately large scale).

An important application of topographic information is to define the hydrological connectivity of basic hydrological computational elements of a model. These elements may be hydrological subbasins or grid elements. The model domain may be a river basin or a set of atmospheric model grid elements. In any case, a set of methods is needed to merge digital terrain, stream location, and existing basin boundary data to establish additional drainage boundaries relative to key locations in the stream channel network and to establish the hydrological connectivity of model elements. The research need is not so much to develop new methods but rather to organize some of the existing methods into a robust and user-friendly system to satisfy many of the needs for basin boundary locations and for hydrological connectivity. (The USGS/WRD and NOAA/NWS are developing a project to address some of these watershed basin and stream network delineation issues, especially standardization of algorithms and data).

The resolution at which stream network data are needed varies depending on the application. Digital stream locations data are available for the entire United States at several resolutions ranging from 1:250,000 to 1:24,000 scale.

OBJECTIVE: Develop strategies to use available topographic information for model development and model parameter estimation, and investigate approaches suitable to obtain required multiresolution topographic data on a global basis.

Activities in support of this objective include:

• Define overall GCIP modeler and scientific investigator requirements for multiresolution topographic data including derivative topographic characteristics, documentation of available data sources, and assessment of data adequacy for GCIP Principal Research Areas.

• Organize existing topographic data analysis tools and algorithms into a user-friendly software package that will facilitate the generation of basin boundary locations and hydrological networks from existing topographic data resources, as well as hydrological modeling research.

• Facilitate hydrological modeling research that is focused on determining which topographic properties, including appropriate horizontal and vertical DEM resolution and accuracies, are essential for properly modeling the effects of hillslope processes on the surface water budget and on the timing of hillslope runoff.

• Determine the adequacy of available multiresolution topographic data sets to meet model requirements based on research results in the preceding activity.

• Investigate remote sensing technology as part of a sampling strategy to develop regionalized hillslope statistics that are suitable for global data set development, especially in other GEWEX project areas.

OVERALL OBJECTIVE: Improve the description and understanding of the radiative fluxes that drive land-atmosphere interactions and their parameterization in predictive models.

6.4.1 Satellite Product Development

OBJECTIVE: Produce satellite products to define spatial and temporal variability of clouds and radiation over the Mississippi basin.

Activities to support this objective include:

• Development of high-resolution radiation products for the LSA-SW or ARM/CART area.

  The components of the Earth's radiation budget at the top of the atmosphere--planetary albedo and outgoing longwave radiation (OLR)--are routinely derived by NOAA/NESDIS from the AVHRR on NOAA's polar orbiters and will be part of the derived data products of GCIP. But polar satellite observations provide only two measurements per day for each area: one in the daytime and one at night. Clearly, for land/atmosphere interactions the diurnal variation of radiation is a key factor, and the geostationary satellites can provide such information.

  Algorithms for deriving planetary albedo and insolation from GOES observations of reflected solar radiation have been developed by several investigators (e.g., Pinker and Laszlo 1992). Further research is needed to accurately retrieve the vertical profile of shortwave and longwave radiative fluxes.

  GOES longwave products [OLR, downward longwave radiation (DLR), and longwave cooling (LC)] can be derived from GOES sounder data using the techniques developed for the polar-orbiting sounder data [the high-resolution infrared sounder (HIRS)] (Lee and Ellingson 1990; Ellingson et al. 1994a; Ellingson et al. 1994b; Shaffer and Ellingson 1990). Although the satellite platforms are quite different (geostationary vs. polar orbiting) with sharply differing altitudes, the structure of the algorithms will be quite similar. The OLR will be estimated from the sounder channels as the weighted sum of radiance observations in a number of narrow spectral intervals. Regression equations relating DLR and LC to cloud-cleared sounder radiances and effective cloud fraction will be derived. Most of the progress to date in satellite OLR, DLR, and LC have been for cloud-free conditions. The difficulty in making radiation budget estimates under cloudy sky conditions is related to problems in determining accurate cloud base altitude from satellite observations.

  The clear sky OLR, DLR, and LC that are obtained from the GOES sounder will be compared with equivalent values derived from the polar sounder for identical targets and for times of observation that are reasonably close.

• Development of high resolution spatial and temporal cloud products for the LSA-SW

  A gridded version of the GOES ASOS cloud cover product will be generated routinely at the NOAA/NESDIS within a few months. This product is derived from the GOES sounder and is produced hourly whenever the sounder data are available. This product is generated without using the visible band of the sounder and is fairly accurate at higher levels in the atmosphere, but it may not do a good job of estimating low level clouds or subpixel cumulus clouds. Both the POES and GOES satellite cloud products will provide cloud information for the GCIP continental-scale area at 0.5° spatial resolution and hourly (GOES) to twice daily (POES) time resolution.

  A significantly more accurate, high resolution satellite cloud product is needed from the GOES (hourly) that would provide cloud information at a resolution of 5 - 10 km. This product would allow validation and improvement in the cloud physics parameterization (and from this, the radiation physics and surface energy fluxes) in climate and NWP models. The imager on the current generation of GOES is a 5 channel instrument with resolution and bands very similar to AVHRR. An automated cloud detection algorithm (CLAVR) has been developed and tested on AVHRR data. This algorithm does a better job of cloud detection than the sounder product described above because CLAVR makes use of the very high contrast between land and clouds in the visible band. To properly meet the GCIP cloud requirements, CLAVR, modified to work with the GOES imager data should be developed and implemented. Such an algorithm applied to GOES imager would be of much use for snow/cloud discrimination applied to snow mapping.

(1) validate the NOAA operationally-based retrievals of radiation and cloud parameters, especially the new product list from the GOES I spacecraft series (described in the previous section ).

(2) regionally validate the fluxes from the GEWEX global-scale Surface Radiation budget (SRB) Project (Whitlock et al. 1995);

(3) foster the development of Satellite and Atmospheric Radiation Budget (SARB) retrievals in the EOS Clouds and the Earth's Radiant Energy System (CERES) (Wielicki and Barkstrom 1991) and in the French-Russian Scanner for Earth Radiation Budget (ScaRab); then validate CERES and ScaRab retrievals of the SARB;ScaRab was launched in February 1994, and it functioned until March 5, 1995. A preliminary comparison of ScaRab with the ERBE wide field of view (WFOV) measurements for March 1994 is favorable (T.D. Bess, personal communication, NASA La RC).

(4) expand the use of ARM, SURFRAD, and BSRN surface-based measurements to operational satellite systems and to the MODIS (Moderate Resolution Imaging Spectrometer), MISR, ASTER (Atmosphere Surface Turbulent Exchange Research facility,CERES, and AIRS (Advanced Infrared Studies) sensors on EOS.

Recent advances in fast radiative transfer techniques (i.e. Fu and Liou 1993), in satellite remote sensing, and in the deployment of surface instruments in the GCIP region permit the development of a more accurate and comprehensive description of the radiative fluxes in the atmospheric column. Previous efforts to obtain radiative fluxes by remote sensing have concentrated on the surface (SRB) and the top of the atmosphere (TOA). The full vertical profile of broadband fluxes, as well as the narrowband radiances observed by the satellites, can now readily be computed and compared with measurements at a number of sites. A more internally consistent description of atmospheric radiation is thereby produced. The resulting surface fluxes can be used to validate the operational retrievals described in the previous Section 1. They also serve to test the satellite-based retrievals of clouds, which are used for the calculations. The within-the-atmosphere flux profiles (SARB) can be used to test the fluxes produced by mesoscale and general circulation models. The SARB is the basic driver of the hydrological cycle, the general circulation, and global change.

Version 1 of the CERES/ARM/GEWEX Experiment (CAGEX) contains such a comprehensive radiative description of the atmosphere in the longwave (LW) and shortwave (SW). CAGEX (Charlock and Alberta 1995) Version 1 provides, for 26 days in April 1994, a space-time grid with:

(a) satellite-based cloud properties, aerosol, and atmospheric sounding data that are sufficient for broadband radiative transfer calculations;

(b) vertical profiles of radiative fluxes calculated with that data as input; and

(c) validating measurements for broadband radiative fluxes and cloud properties.

CAGEX is available by anonymous FTP: ( http://www.arm.gov/docs/data/CAGEX.html, with instructions). Version 0 was issued in February 1995 at NASA Langley, where it was used to test the Gupta LW algorithm for the next phase of the GEWEX SRB Project. CAGEX is used to test radiation codes at GKSS (Germany), McGill University (Canada), ECMWF, and other institutions. Version 1 also has SW fluxes and aerosol data. Version 2.0.0 of CAGEX covers the ARM Enhanced Shortwave Experiment (ARESE) from Sept. 25 to Nov. 1, 1995. New features of Version 2.0.0 include (a) multiple sets of sounding data from 3-hourly ARM radiosondes, from instruments like the ground-based AERI (LW spectrometer), the MWR (microwave) and the GPS receiver, and from the NCEP Eta model output, (b) broadband surface radiation measurements from RAMS (Valero et al.) and adjustments to standard observations based on cavity measurements (Michalsky et al.), (c) vertical profiles of aerosol from the MicroPulse Lidar (MPL; Spinhirne and Hlavka), (d) cloud profiling radar data (Clothiaux), (e) cloud LWP from MWR, (f) changes to the Fu-Liou code including the insertion (Kratz and Rose) of the CKD (Clough et al.) LW H2O continuum, (g) calculations with aerosol optical properties for various mineral dust particle sizes (Tegen and Lacis), in addition to the original d'Almeida et al. aerosols, and (h) modifications to the Minnis et al. GOES-8 TOA fluxes and cloud property retrievals.

The clear-sky data in CAGEX Version 2.0.0 has been designed, as per the "Open SW Workshop" at the AMS Ninth Conference on Atmospheric Radiation at Long Beach (Feb. 97), to permit more rigorous testing of SW radiative transfer routines and input data, as well as measured fluxes. Clouds observations in Version 2.0.0 have more redundancy; optical depth from GOES-8 and LWP from surface MWR; height of cloud top from GOES-8 and radar; measurements of the height of cloud base from lidar and radar (and estimates from GOES-8).

In CAGEX Version 1's for April 1994, the computed SW insolation for clear skies generally exceeded the observations; the discrepancy for cloudy skies was similar. By using different aerosol optical properties, some colleagues (Trishchenko, Li, Fu and others) have reduced or eliminated the clear sky discrepancy for April 1994. The clear sky discrepancy for Version 2.0.0 (Fall 1995) appears to be more robust. A large cloudy sky discrepancy for ARESE October 30, 1995 (first reported by Pope and Cess using aircraft data) is quite apparent when comparing computed fluxes with satellite and surface data in Version 2.0.0.

CAGEX Version 2.0.0 has been used to test the vertical profiles of humidity, SW diabatic heating, and LW diabatic cooling in the NCEP Eta model, which activated a prognostic scheme for clouds during fall 1995.

One surprising result in CAGEX is the demonstration of a significant discrepancy between measured and computed SW fluxes at the surface for clear skies; this has been confirmed by various ARM researchers in ARESE. In the NASA EOS, CAGEX serves as a window for community-wide access to preliminary retrievals of fluxes and cloud properties in the CERES program. CAGEX fluxes are determined with the Fu and Liou (1993) delta-4-stream radiative transfer code using the Minnis et al. (1993) cloud retrievals. Experiments with tuned fluxes, in which atmospheric constituents are adjusted to cause computed and observed fluxes to better match, are underway ( Charlock et al. 1994). For limited time periods, within-the-atmosphere fluxes as measured by Unmanned Aerospace Vehicles (UAV) will be inserted in the data stream. Subsequent versions of CAGEX will be used to validate CERES determinations of atmospheric fluxes and similar exercises using ISCCP and ScaRab. Hence CAGEX will continue well after the launch of CERES on TRMM in November, 1997 and EOS-AM (1998). The MODIS and CERES teams in EOS are now drafting plans for a concentrated validation effort over the ARM/CART site in September 1998. "Joint validation planning among the MOPITT, MISR, ASTER, CERES, MODIS, and SAGE III teams were discussed at the Workshop on Atmospheric Validation in EOS-AM1 and SAGE III (WAVES) at Hampton University in October, 1997.

The dense coverage of measurements over the ARM site are presently supplemented with the geographically dispersed SURFRAD described later in this section. When combined with comprehensive satellite-based retrievals and radiative transfer calculations, SURFRAD will provide a rigorous measure of the radiative forcing of climate at selected sites. For example, the present satellite-based record of the interannual variability (IAV) of snow cover lacks an exacting validation in terms of radiative flux; this poses a great uncertainty in monitoring a key climate feedback. There is a corresponding uncertainty in radiative forcing of aerosols; measurements of aerosols and measurements of fluxes have not been matched with calculations to satisfactory accuracy. The SURFRAD monitoring sites at Fort Peck, Montana (high seasonal snow cover and IAV) and Bondville, Illinois (large annual loading of atmospheric sulfur) are well-suited for diagnosing the impacts of snow and aerosols when combined with calculations such as CAGEX (above) or with the NOAA retrievals (Section 6.5.1), which are based on operational satellite data.

The procedures honed in these exercises will be used again with more advanced MODIS, MISR, ASTER, and CERES sensors after the launch of EOS-AM in 1998. In preparation for CERES, helicopter measurements of the SW bidirectional reflectance function (BDRF with about 10 nanometer resolution spanning the shortwave spectrum, the LW window directional radiance, and the broadband SW and LW fluxes (i.e., Purgold et al. 1994) are planned for the ARM/CART site during the spring of 1998. The helicopter measurements are vital for improving the integration of space-based and surface-based data for two reasons. First, they are needed to determine the full angular dependence of surface radiation; a given satellite measurement covers only a single angle. Second, they are needed to determine the spatial distribution of radiation about the surface radiometer; the surface radiometer covers only a tiny area. It is hoped that resources will permit helicopter measurements over some SURFRAD and BSRN sites, too. Another supplement to routine surface measurement is enhancement with a spatial network of instruments. In conjunction with CERES preparations during the fall of 1995, NASA Langley deployed a network of five additional radiometer sites to supplement CAGEX retrievals of surface fluxes in the ARM Enhanced Shortwave Experiment (ARESE). The enhanced spatial network measures fluxes over a large area, as does a satellite pixel, permitting a more realistic validation of the satellite results.

The combination of (1) detailed radiative transfer calculations, (2) satellite-based retrievals, and (3) surface measurements as anticipated in GCIP will permit a significant advance in the description of atmospheric radiation and associated forcings and feedbacks. Supplements to the surface measurements are needed, however; only a single helicopter survey of ARM is definitely planned; deployment of photometers and cloud lidars at more surface sites is uncertain; the determination of aerosol optical properties is a step forward but not the answer; and snow sites especially should have a network of radiometers on towers.

6.4.3 Validation and Improvement of Operational GOES Shortwave Radiation Budget Products

The operational production of downwelling and upwelling shortwave (SW) and photosynthetically active radiation (PAR) for GCIP is done using the University of Maryland algorithm (Pinker and Laszlo 1992), as modified for the GOES 8/9 imager. The model also allows estimation of top of the atmosphere shortwave radiative fluxes. The procedure uses clear sky and cloudy top of the atmosphere calibrated radiances in the visible band, the cloud fraction in the target, and information on the state of the atmosphere, as available in real-time from the Eta model, as input to the algorithm. Snow information is also appended, as available fa new routine snow map product at NESDIS. Cloud detection is done with a two threshold method, from visible data only. The new GOES 8/9 procedures, namely, the algorithm, the cloud detection methods, the atmospheric input parameters, and changes in calibration, need to be evaluated. The need for incorporation of seasonal/monthly surface type models in the shortwave algorithm has also to be evaluated.

A process has been established whereby the University of Maryland accesses the GCIP insolation products as generated at NESDIS, as well as the input files used at NESDIS to generate the product. The input files are used to run the model off-line, compare with the product produced at NESDIS, and to test various options in the model configuration. Of particular interest are possibilities to optimize the models operation and/or introduce simplifications. The model output will be validated against ground observations,to include, in the near future, observations from SURFRAD, BSRN and ARM/CART. Ground truth data for PAR are also needed for validation of this component of the SRB. This process is essential for achieving the best possible accuracy from satellite products.

In addition to the NOAA GOES-8/9 and POES operationally based retrievals in GCIP, the NASA CERES is sponsoring a more limited domain program of research retrievals of the SARB (Charlock et al. 1994). Satellite-based cloud retrievals, meteorological data, and radiative transfer calculations will be used to retrieve the SARB over the ARM/CART site in Oklahoma. Computed fluxes and radiances will be compared with ARM-observed surface and unmanned aerospace vehicles (UAV) fluxes, as well as with other satellite data. Tuning algorithms will subsequently adjust atmospheric and surface input parameters, bringing the calculated SARB to closer agreement with observations. Results of the SARB retrievals will be compared with those of other groups and with data. The aim is to develop accurate retrievals of the SARB based on satellite data and to foster the development of such retrievals in the atmospheric sciences community. The first research data set in this CERES/ARM/GEWEX activity covers the April 1994 IOP. In a 3 x 3 matrix with 0.3° increments, daylight cloud retrievals every 30 minutes are provided from GOES-7 with the Minnis et al. (1993) cloud retrievals for cloud albedo, cloud center height, cloud amount, cloud center temperature, cloud thickness, cloud infrared (IR) emissivity, cloud reflectance, cloud optical depth, cloud top height, cloud IR optical depth, cloud mean IR temperature, and cloud top temperature. In a subsequent ARM IOP, Dr. Charles Whitlock plans to employ a helicopter to measure the spectral bidirectional reflectance of the surface. This measurement will permit a detailed study of the clear as well as cloudy sky effects of the surface and aerosols on the profile of radiative fluxes.

The SARB drives the hydrological cycle, the general circulation, and the global climate change. The SARB computed by GCMs is not regarded to be sufficiently reliable for accurate climate prediction. The state of numerical weather prediction (NWP) model simulations of the SARB limits medium-range weather prediction, too. We lack an adequate observational record of the SARB either in clear or cloudy skies. Cloud feedback is generally considered vital to climate but remains uncertain. More fundamentally, forcing occurs, as well as feedback uncertainties because of the radiative effects due to atmospheric aerosols and the Earth's surface.

An observational SARB record is needed for the validation of GCMs and for diagnostic investigations of low-frequency variability and secular climate change. The development of an observational record of the SARB is one objective of the CERES activity ( Wielicki and Barkstrom 1991) in the EOS and GEWEX. The array of instruments deployed by ARM over the CART site presents a unique opportunity for developing and validating satellite-based retrievals of the SARB. The ARM/CART site is well suited to observing the profile of atmospheric water vapor, the vertical and horizontal structure of clouds, and aerosols; these parameters, as well as the ARM/CART surface and UAV measurements of radiometric fluxes, are critical for testing satellite-based retrievals of the SARB. Activities to support this objective include:

• Retrieval of surface and atmospheric radiation budgets from satellite and meteorological data over the ARM/CART site.

• Comparison of computed fluxes and radiances with ARM-observed fluxes and other satellite data, and NWP model outputs.

• Development of techniques to retrieve aerosol and land-surface radiative forcing with satellite and ground-based measurements.

A gridded version of the Automated Surface Observing System (ASOS) clouds will be generated for GCIP as a continental-scale product. The ASOS clouds are produced operationally from GOES at weather station locations to supplement the laser ceilometer observations of the ASOS of the modernized weather service. The ASOS clouds are generated from the GOES sounder using the carbon dioxide slicing technique ( Menzel and Strabala 1989; Wylie and Menzel 1989). They can also be generated from the image data by substituting the water vapor channel for the carbon dioxide band. Whether the sounder or imager version is implemented depends on which technique is chosen by the NWS for the operational ASOS product. In addition to cloud information, the ASOS-cloud processing system produces clear sky surface temperature as an intermediate product, which will be evaluated for surface energy budget studies and validation of the Eta and other models.

CLAVR stands for clouds from the advanced very high resolution radiometer (AVHRR) on the POES. NESDIS has developed this cloud product over the last few years, and it is currently being generated on a routine basis from the afternoon POES observations (Stowe et al., 1991). This product includes cloud amount, type, and height of each cloud type at a resolution of one degree in latitude. During GCIP it will be produced routinely on a global basis by NESDIS for day and night from both POES spacecraft. The NESDIS will access the product to produce a CONUS sector for the GCIP database.

The ASOS cloud product produced from the GOES data meets the needs of GCIP users better than the CLAVR cloud product produced from POES data. We shall therefore select the ASOS product as the "best available now" for GCIP with the CLAVR to be used in the event of difficulties with the ASOS product.

6.4.6 Radiation Data Products

Radiation data sets are required for the GCIP EOP on a continental scale. This information will include top-of-the-atmosphere, surface, and atmospheric radiation data based on both POES and GOES observations.

6.4.7 Outgoing Longwave Radiation (OLR) and Planetary Albedo

The OLR and planetary albedo radiation budget products have been obtained from multispectral, narrowband radiometric scanners for many years. This product is currently being produced using a technique to infer the OLR from four of the channels on the high-resolution infrared sounder (HIRS) flown on the POES(Ellingson et al. 1989; Ellingson et al. 1994a).

The above methodologies for obtaining top-of-the-atmosphere, OLR, and planetary albedo are being applied to GOES-8 data and are being produced for GCIP.

6.4.8 Surface and Atmospheric Radiation Budget Components

In addition to the OLR, methods have been developed to infer the downward longwave radiation (DLR) flux at the surface (Lee and Ellingson 1990) and the vertical profile of longwave cooling (LC) (Shaffer and Ellingson 1990; Ellingson et al. 1994b) from POES observations. The DLR and LC estimation techniques require spectral radiance data from the HIRS and the vertical distribution of cloud amount and cloud base height. The NESDIS is implementing the techniques in an experimental operations test in the TOVS sounding system.

Insolation and photosynthetically active radiation (PAR) for the GCIP CSA (and in fact, for the whole U.S.) will be produced from GOES 8/9 imager observations. The insolation algorithm, developed at the University of Maryland (Pinker and Ewing 1985; Pinker and Laszlo 1992) is a physical algorithm that uses GOES imager observations of reflected visible radiation. The algorithm uses target clear radiance, target cloudy radiance, fraction of clouds in the target and atmospheric precipitable water (from the Eta model). Other required input to the model is surface albedo (Matthews 1985) and snow cover. Net solar irradiance at the surface can be derived from the insolation and surface albedo.

This algorithm has been modified at the University of Maryland to use GOES 8/9 data as input. A two threshold cloud detection method has been developed that provides the clear and cloudy radiances and the fractional cloud cover required by the algorithm. Over the past two years the insolation algorithm has been implemented into the GOES sounding system at NESDIS and routine production has begun. The products are not operational, however, but are currently experimental and generated specifically for GCIP.

Because the insolation algorithm is newly developed for GOES 8/9 data, it is vital that the insolation estimates be compared with ground truth and all aspects of the procedure, from cloud detection through insolation production, and be subject to modification and improvement. This way, the accuracy and reliability of the products will increase, thereby meeting one of the main objectives of GCIP.

Outgoing longwave radiation, DLR at the surface, and atmospheric LC rates will be derived from GOES-8 by applying the methodologies used to generate these quantities from POES-HIRS observations. Some development is needed to apply the techniques to GOES data.

In the case of clear skies, surface temperature will be retrieved from the GOES shortwave radiation budget processing. For the clear radiances for each target a split-window surface temperature will be applied. At first simple algorithms that assume a unit surface emissivity will be used, but research is needed to develop an algorithm that adjusts for the different surface emissivity of a variety of surface types. Estimates of surface temperature can be used to obtain upward longwave radiation fields at the surface. It is also important that land surface temperature be retrieved where it is cloudy by use of microwave (AMSU) window channel data. Such products are being developed at NESDIS for NOAA K-M, and will be available to GCIP.

There is another source of surface temperature that should be considered for GCIP. This is the Derived Product Imagery (DPI) which includes surface skin temperature, lifted index, and total precipitable water. The DPI is a planned operational suite of products from the GOES 8/9 imager that is currently under active development. The resolution of the surface temperature in the DPI is 4 km, so in addition to averages of surface temperature for targets of about 50 km. resolution, histograms of surface temperature could be saved. This could be of considerable interest to the modeling community.

6.4.9 SURFRAD Sites for GCIP

Six Surface Radiation (SURFRAD) sites are planned for the contiguous 48 states (three of these are already installed in the Mississippi River basin). This network is intended to provide high quality, long term solar and infrared radiation measurements for a variety of research needs: to validate satellite-derived surface insolation; to provide a long term climatology of surface radiation measurements (at least 25 years); to detect trends in surface radiation; and, to verify radiative transfer models. The basic instrumentation set (see Table 6-1) includes radiometers for upwelling and downwelling solar and INFRARED radiation, a sun-tracking normal incident pyrheliometer (NIP) for measuring direct solar irradiance, and a meteorological tower. Other special sensors may be added.

Table 6-1. Basic Instrumentation at a Surfrad Site.

MEASUREMENT	NAME	COST ($)	ACCURACY
Direct Solar Irradiance	Cavity radiometer (required at BSRN) shadow band radiometer NIP	18,000 10,000 1,800	2 Wm-2 5 Wm-2
Diffuse Solar	Pyranometer (2(pi) solar flux) (radiation >2.5 pm filtered out)	1,800	5 Wm-2
Global Solar (direct and diffuse)	Pyranometer (no tracker)	1,800	10 Wm-2
Reflected Shortwave	Inverted pyranometer (shaded from sun)	2,000	10 Wm-2
Downward Longwave	Pyrgeometer (filtered pyranometer)	2,850	6-8 Wm-2
Upward Longwave	Inverted Pyrgeometer	2,850	6-8 Wm-2
Photsynthetically Active Radiation	PAR Instrument (filtered silicon detector)	200	TBD
Surface Meteorology Tower	10-m height: winds, pressure, temperature, humidity	6,000	TBD

The URL address: ( http://www.srrb.noaa.gov) has detailed information on SURFRAD sites, instrumentation, and access to data. In addition to the instrumentation mentioned in Table 6-1, NOAA has obtained Multi-Filter Rotating Shawdowband Radiometers (MFRSR) for SURFRAD. Operational MFRSR algorithms retrieve column aerosol optical depth, predictable water, and ozone; research algorithms provide cloud optical depth. The SURFRAD combination of broadband and MFRSR measurements will permit the estimation of aerosol direct radiative forcing to climate over GCIP.

SURFRAD sites have been chosen to be representative of extended regions. Each has reasonably uniform and stable surface properties that are representative of the region. This requirement is the primary concern of those doing verification of satellite-based algorithms. Those who will use SURFRAD data to verify the satellite-derived surface radiation data require that the area surrounding the sites be spatially uniform over at least the area of one GOES-8 sounder pixel, which is 10 km (E-W) by 40 km (N-S).

One SURFRAD site in the GCIP region is at Bondville, Illinois, located approximately eight miles southwest of Champaign, Illinois. It is owned by the University of Illinois Electrical Engineering Department and managed by the Illinois State Water Survey. This site consists of six acres of grassland (being updated to 14 acres) and surrounded by 220 acres of soybeans and corn. This site is currently operational and also contains a suite of aerosol measurement systems operating under a separate NOAA funded aerosol monitoring program. A second SURFRAD site in the GCIP region is the Poplar River site (near Fort Peck, Montana). The Poplar River flows south out of Canada and into the Missouri River. This site has good hydrological data available and the Poplar River is not used for irrigation (because of high levels of alkali). The site is on rangeland with no trees in northeastern Montana. This site was operational in the summer of 1994. A third SURFRAD site in the GCIP region is the Goodwin Creek site (near Oxford Mississippi). The Goodwin Creek Experimental Watershed is an ARS site located in northern Mississippi. It is relatively flat, and its land use is about 14 percent agricultural, 26 percent timber, and 60 percent idle pasture land. Four lakes are in the region. This site is operational since the fall of 1994. The data from these sites will be quality controlled by NOAA's Air Resource Laboratory (ARL) in Boulder, Colorado. Data will be archived at the ARL facility in Oak Ridge, Tennessee and accessible via the GCIP in situ data source module.

In addition to the usual radiation and hydrological measurements at the three SURFRAD sites identified earlier, funds have been requested to add instrumentation for the following: soil moisture, snowfall measurements (in the northern sites), ground heat flux, and cloud determination via lidar and/or possibly digitized pictures.

Not all the requested instrumentation will be immediately available at all the GCIP SURFRAD sites. It is expected that further implementation of instrumentation will likely occur as more resources become available and become part of the normal operations at the three SURFRAD sites.

6.5 Streamflow and Runoff

Streamflow data and runoff estimates are required both for the development and for the testing and verification of coupled atmospheric/hydrological models. Testing and verification may be approached in two complementary ways. First, runoff from the coupled models can be verified by routing the runoff from a number of grid points (10 or more) to a streamgauge and comparing the model discharge with the observed discharge on a designated basis. The gauges used for this purpose must be essentially unaffected by upstream regulation or diversion. In practice, most of the continental discharge gauges are influenced by regulation and diversion, and may not be good choices for verification (except perhaps on an annual or climatological basis). Therefore, a second complementary approach to compensate for these upstream effects is needed.

6.5.1 Relationships between Runoff and Streamflow

OBJECTIVE: Develop and apply improved techniques for the determination/estimation of runoff and streamflow appropriate to the scales of primary interest to GCIP.

Activities to support this objective follow:

• Development of globally applicable routing models appropriate for the scale of atmospheric models.

• The runoff routing problem has two components. The first is to account for the time delay for water to flow over and through hillslopes and the ground water system to the stream network and to pass through the upper, highly disperse reaches of the stream network. This time lag is often accounted for in hydrology using a "unit hydrograph." Globally transferrable applicable synthetic unit hydrograph approaches or some mathematically equivalent alternatives must be developed and tested, including nonlinear alternatives.

• The second component of the routing problem is to account for the time that water flows from upstream gauges to those downstream or from the runoff generated from the atmospheric model grid through intervening grids to a streamflow measuring point downstream in the river network. Although the equations describing the unsteady flow of water in river channels are well known, further work is needed on methods of estimating a priori parameter values to apply these equations to specific river reaches globally. This estimation could include developing and testing various simplified, globally applicable approaches to the solution of the full unsteady flow equations using geographic information systems to estimate channel slope and other hydraulic parameters. These approaches must handle "leaky rivers" and account for the natural losses in rivers and marsh areas.While routing may not be critical for estimating water budgets over a month this may not be the case for extremes and routing effects cannot always be ignored.

• Identification of tributaries of the Mississippi River basin and/or periods where water management effects can be neglected, and the subsequent evaluation of runoff predictions from atmospheric models by routing to a streamgauge and comparing the data with observed streamflow.

• Improved understanding and description of the effects of hillslopes and stream channel nonlinearities on the amount and timing of downstream discharge for large continental catchments and major tributaries thereof (drainage areas typically greater than 5000 km²).

• Evaluation of the error in methods for estimating gridded runoff fields as a function of catchment area and aggregation period using streamflow measurements independent of those used to estimate the runoff grid.

• Gridded monthly runoff data (on a 30-minute grid) are needed to assess the coupled model validation and diagnostic aspects. The model representation of the surface water budget depends on both local and large-scale processes. To understand how to improve the limitations indicated in model variations at specific streamgauges, additional information on a larger spatial scale is fundamental. For an initial comparison of coupled model gridded runoff, reconstituted runoff that accounts for diversions and ignores reservoir operations would be the simplest approach to developing diagnostic contours of runoff. This approach would enable a more qualitative comparison of the spatial variability of actual and model runoff and would enable the researcher to look more clearly at the various parts of the water budget. Distribution functions could be developed to obtain a better space-time resolution of the water budget components. The emphasis would be on the distribution, not necessarily the actual numbers. If reservoir storage effects are significant, they should be taken into account in developing the reconstituted flows.

• The grid-mapping approach of river discharge was recently reviewed by Arnell (1995). Five methods are considered. These methods and other appropriate approaches need to be evaluated in relation to related activities of agencies in the Mississippi River basin. The Global Runoff Data Centre (GRDC) is coordinating the data for a German-funded project, "Transformation of measured flow data to grid points" as a contribution to World Climate Programme (WCP)-Water Project B.3. The pilot area under study covers the basins of the Rhine, Weser, Elbe, Oder, and Weichsel Rivers within Germany, Czechoslovakia, and Poland. The results of this project and further work with European data by the UK Institute of Hydrology will assist in planning the best approach for the Mississippi River basin.

• Development of algorithms to estimate the uncertainty in gridded runoff fields as a function of drainage network configuration, streamgauge location, and space-time scale for cases with minimal water management effects.

• Development of improved methods for better estimating gridded runoff by evaluating the relative contributions of space-time aggregation, channel network and gauge configuration, and water management effects on the error in gridded runoff fields.

The above activities will be supported by the following specific activities and outputs in 1998-2000.

• Extend the available historical data base for unregulated basins at the SSA and ISA scales (10 to 1000 km²) in the LSA-SW (Arkansas-Red River basin) by updating from 1988 the active streamflow stations on the Wallis-Lettenmaier-Wood CD-ROM and the USGS HCDN CD-ROM. Include additional from the archival record that have shorter periods of records than those on the existing data sets, e.g., by adding two additional categories of unregulated stations which have 10 and 20 years of data. The purpose would be to develop and demonstrate regionalization methods for the estimation of hydrologic model parameters. In addition to allowing the estimation of the land-surface model parameters these data are needed for the development of runoff routing parameters and gridding runoff. This work to quality control and fill in missing data is being undertaken by the University of Washington. Extension to the LSA-NC could be included as part of the NWS WARFS initiative and the work within the NWS/NESDIS Core Project to develop the required historical data bases.

• Develop naturalized streamflow records at key locations in the LSA-SW up to the current time to enable the validation of the atmospheric model predictions. Key locations would include the Red River at Shreveport and the Arkansas River at Little Rock, being the largest basins which can be feasibly considered. This will require agency interaction, particularly with the COE, and the updating of calculated flows by the COE, or the acquisition of the reservoir storage algorithms and the algorithms used in the reservoir operating rules.

• Test a method for estimating gridded runoff data for the LSA-SW to enable the direct validation of atmospheric model runoff predictions. This activity will require research funds and may be supported by the NCGIA, working in conjunction with the USGS and the UK TIGER project supporting GCIP.

As an alternative to naturalized flows, compute the regulated runoff from atmospheric models by using runoff routing and reservoir storage models. The model feasibility has already been demonstrated. Model parameters from the NWS ABRFC are already available, together with their conversion to the application of gridded or distributed models as part of the NWS/NESDIS core Project and the macro-scale model parameters developed over the Arkansas-Red River basin by the University of Washington, the models and parameters will be available in 1996.

6.5.3 Surface and Ground Water Measurements

The primary observations of hydrological variables are from in situ networks and consist of stream gauges, measuring wells, measurements of water storage in large reservoirs, soil moisture, evaporation and estimates of snow cover. GCIP is treating soil moisture as a separate variable (see Section 6.2) and also estimates of snow cover. (see Section 6.1). There are few measurements of evaporation available. This leaves stream gauges, measuring wells and measurements of water storage which are needed to provide derived information for computing water budgets. In cooperation with many other Federal, state, and local agencies, the USGS collects water data at thousands of locations throughout the nation and prepares records of stream discharge (flow), and storage in reservoirs and lakes, ground-water levels, well and spring discharge and the quality of surface and ground water. The number of stations collecting such data was summarized in Table 1 of the GCIP Implementation Plan, Volume I (IGPO 1993), and is updated for each of the data sets compiled by GCIP.

Most of the gauged streams in the Mississippi River basin are affected by various water management activities such as upstream storage and diversion for human activities and irrigation. The USGS has a hydrological benchmark network of 58 stations virtually unaffected by human activity distributed across the United States (Lawrence 1987). Wallis et al. (1991) prepared a set of 1009 USGS streamflow stations for which long-term (1948-88) observations have been assembled into a consistent daily database and missing observations estimated using a simple "closest station" prorating rule. Estimated values for missing data, as well as suspicious observations, are flagged. The data are retrievable by station list, state, latitude-longitude range, and hydrologic unit code from a CD-ROM. This data set is being updated to include the years since 1988 with primary emphasis on those stations important to GCIP. Landwehr and Slack (1992) compiled measured streamflow data for 1659 stations with at least 20 years of complete records between 1874 and 1988.This data set is available by anonymous ftp at URL: ftp://ftprvares.er.usgs.gov/hcdn92. This data set is also available on CD-ROM and is being updated with post-1988 data. A streamflow data product similar to those described above will be produced for the GCIP EOP.