"The LEWS technology package has demonstrated that satellite weather and NDVI data can be incorporated into an analytical system to simulate forage conditions over large regions, map those conditions, and forecast likely responses."

• Introduction
• LEWS technology, information and analysis design
• Establishing foundation weather data
• Establishing vegetation monitoring sites
• Forage deviation indexing
• Using geostatistics to map forage conditions
• Forecasting 90-day forage conditions
• Linking LEWS with decision makers
• Passing the "Believability Test"
• Communicating warnings
• Summary and conclusions
• Acknowledgments
• References
• Author information
• Additional web resources

Introduction

(Back to top)
The delicate balance between selecting and maintaining a stocking rate that meets the short-term economic goals needed for ranch or pastoral household survival versus one that sustains long-term livestock carrying capacity has long dominated the decision making process of livestock producers worldwide. This process is driven by the level of human needs of the decision maker in relation to the level of risk an individual is willing to undertake given his/her current livelihood. Scientists in natural resources management and meteorology have attempted to develop technologies to help mitigate these risks by providing information that allows more timely decision-making. Until recently, these technologies have relied on use of flexible stocking strategies for different kinds and classes of animals, reserve grazing areas, rainfall deviation from average, or a series of indexes that have been derived from crop production systems; their low adoption rates within the livestock industry reflect their limited success in supporting the stocking decision process.

Over the past 10 years, the Ranching Systems Group (RSG) within the Center for Natural Resource Information Technology (CNRIT) and the Center for Grazinglands and Ranch Management (CGRM) at Texas A&M University have been working on a new suite of technologies that help mitigate the risk of drought and improve adoption rates of the information that is released. The overall technology described herein was initially funded in 1997 by the United States Agency for International Development (USAID) Global Livestock Collaborative Research Support Program (GLCRSP) administered by the University of California-Davis for development of a Livestock Early Warning System (LEWS) in East Africa. The technology has subsequently been configured for use in the United States as part of a LEWS pilot program in Texas and has served as the basis for design of a new National Range and Forage Loss Insurance program administered by USDA-Risk Management Agency. This paper describes how the LEWS technology works and the emerging issues associated with adoption of information generated for improving the stocking decision making process. Cross comparisons between the system being implemented for the US ranching industry and technology implemented for East African pastoralists will be highlighted.

Pastoralists in Africa are facing conditions where traditional drought-coping norms are collapsing due to increased population pressure, erratic climatic patterns with higher frequency of drought, limited marketing opportunities, changing land tenure patterns, rising social conflict, limited water supply and greater incidences of disease transmission. In the US, the ranching industry is facing a major change in land ownership patterns and a shifting experience base of managers where "norms of production" are not well established. Profit margins are very low, resulting in a greater focus on risk management tools to mitigate the impact of unexpected events on survival of the ranch firm.

LEWS technology, information and analysis design

(Back to top)
The Phytomass Growth Model (PHYGROW) is the foundation technology used in LEWS for monitoring the impact of emerging weather events on livestock forage supply both in Texas and in the pastoral regions of east Africa. Primary inputs for the model include soil parameters, plant community characteristics, and livestock management decision rules which are driven by gridded weather data for a particular location to simulate daily forage available for livestock and wildlife.

Throughout the development of the LEWS technology, we have attempted to determine what kind of information would be valued by livestock producers. Surveys of ranchers in Texas and pastoral communities in Ethiopia, Kenya, Tanzania and Uganda all revealed the same trend for information needs. Essentially, to support decisions on destocking/restocking or moving to other grazing areas, there needs to be an indication of how the plants translate weather into forage and a quantification of how current conditions deviate from historical average responses. Additional information on what forage conditions were like at the same time last year and a few years back, coupled with trend information and a near-term forecast (preferably 90 days ahead), were also deemed important for improved decisions. Finally, all decision-makers indicated that analyses must be believable, requiring a period where they can ground-truth the information provided on forage conditions. This process builds their confidence in using the information for decision-making (Jochec et al. 2001).

Link to Fig. 1, LEWS system diagram (~14K)

Given the variety of decision environments and governmental support infrastructures that were to be serviced by LEWS, our research team needed to design an automated system to acquire on-line weather data, run the forage production simulation model, and then package the output into a series of products delivered to a wide variety of information users. To overcome these design constraints, a central high-speed server at CNRIT was linked with a series of inexpensive computers using load management software to allow a large number of simulations to be run on a frequent basis. The central server was then linked with the Internet and the model output configured as tables, graphs and maps for near real-time access. In East Africa, we transmitted data to institutions and non-governmental organizations (NGOs) by using new, satellite radio technology; this involved transferring subsets of the data provided on the Internet to a computer server up link site managed by the African Learning Channel in South Africa for broadcasting over the WorldSpace satellite radio system. People on the ground could then link their computers to the satellite radios at a pre-scheduled time and download the LEWS information in approximately 15 minutes, at least three times a week. The information transmitted could be viewed with a standard web browser and/or printed for distribution to key communication nodes in the African LEWS region.

Establishing foundation weather data

(Back to top)
The LEWS technology suite design is similar in both the US and Africa, except for the type of weather data used and resolution of the satellite imagery available. The primary approach of the system is first to establish the boundary of the region of interest. In this case, we were interested in the pastoral regions of East Africa and the grazinglands of Texas. We then established an automated protocol to acquire regional weather data from NOAA weather data web sites and place the data on the CNRIT server. For the Africa system, LEWS uses European Union METEOSAT weather satellite system data that have been processed by NOAA using their Rainfall Estimation (RFE) algorithm for estimating daily rainfall on a 11 x 11 km grid (Herman et al. 1997, Xie and Arkin 1997). CNRIT acquires the African rainfall and NOAA's surface temperature (both maximum and minimum) for the entire continent of Africa, generates solar radiation data, and then makes the data available in model-ready form on the Africa Weather Data web site. For the Texas system, CNRIT acquires the same weather data variables in a 25 x 25 km gridded format for the USA (Higgins, Shi and Yarosh 2000) and makes the data available in model ready form on the US Forage Conditions Weather Data web site.

Establishing vegetation monitoring sites

(Back to top)
The next step involves identifying vegetation monitoring sites which are spatially diverse and represent the major plant communities across the region. Once identified, a representative stand is selected and sampled for grass basal cover (%), forb frequency (% in a 5 x 5 cm frame) and effective woody plant canopy cover (%). It takes approximately 1.5 to 2.5 hours to complete these measurements. The soil type is noted; typical soil series profiles from USDA-NRCS' STATSGO, SURGO and NASIS soils databases are used rather than site-specific soil profiles given the resolution of the weather and satellite imagery data available. When soil attribute data are not available, soil parameters are estimated using the online Soil Water Characteristics: Hydraulic Properties calculator. Since the site being modeled is normally grazed, the rancher or pastoral community is interviewed to determine the typical stocking density and movement/sales rules for the area represented by the plant community(s). In this process, a focus group of producers provides information on when they normally expect to move livestock and identify forage conditions that would force earlier or later movement or destocking, as well as conditions that would allow herd expansion and increased stocking density. Using a panel of grazingland scientists, species in the representative plant communities are then assigned grazing preference values (preferred, desirable, undesirable, emergency, toxic and non-consumed) according to growth phase (rapid, declining, quiescent, dead) for each herbivore to be represented in the analysis, including wildlife species. We have also involved herders on a "bush walk" with an accompanying taxonomist and derived excellent information on differential preferences for plant species by different kinds of livestock.

The weather, plants, soils and grazing rules data are input into a plant growth simulation model called the Phytomass Growth Model (PHYGROW) to generate forage responses for each unique combination. PHYGROW is a hydrologic-based model capable of calculating forage response and hydrologic process of complex plant communities grazed by multiple animal populations. Before PHYGROW simulations can be completed, each plant species must be assigned a series of growth attributes, such as radiation use efficiency, maximum leaf area index, growth temperatures (minimum, optimum and suppression), green and dead biomass turnover rates, and water adhesion coefficients. This password-protected database stores all the LEWS project information on plants, weather, soils, and grazers. PHYGROW data is accessible to our collaborators via the Internet to allow work within the projects or to create new LEWS projects.

Link to Fig. 2, East Africa sites (~36K)

Link to Fig. 3, Texas sites (~21K)

Currently, over 300 monitoring sites across East Africa rangelands are set up for forage production estimation by PHYGROW. These sites are strategically placed to allow limited "sampling" of forage dynamics through use of the biophysical model. Sites are then linked to spatially rich satellite imagery using geostatistical techniques to map regional responses. In the USA, we use the concept of "virtual landscapes" represented inside the 25 x 25 km weather grids. A virtual landscape can be described as a representative plant community and soil combination that is dominant in a weather grid. In the Texas LEWS, there are currently over two hundred 25 x 25 km weather grids established in seven pilot regions of the state, with each grid averaging five different virtual landscapes. The simulations can be set up to allow the user to select a particular plant community or a weighted response given the composition of plant communities in the grid.

Forage deviation indexing

(Back to top)
The foundation analysis of the system involves computation of historical grazed standing crops (kg/ha), percent deviation of current grazed standing crop, relative to the historical daily average grazed standing crop, and the percentile ranking of the current standing crop relative to the historical values on that day of the year. We were fortunate in the USA that NOAA has developed a historical 25 x 25 km gridded weather dataset that starts in 1948 and is updated daily across the entire USA (Higgins , Shi and Yarosh 2000). In Africa, two approaches were used. Initially, LEWS developed a monthly climatic surface of average temperature and rainfall for the East African countries using a splining technique (Hutchinson 1991) from historical weather station data corrected for elevation and proximity to large bodies of water (Corbett et al. 1998). These monthly averages were then used as coefficients in the WXGEN weather generator (Nicks, Richardson and Williams 1990) parameterized for the nearest World Meteorological Organization (WMO) reporting station. This procedure allowed the generation of a 30-year weather data set for each site. This technique was used until 2002. Recently, Funk et al. (2003) combined daily rainfall reanalysis fields, monthly interpolated rainfall, and an orographic precipitation model to derive the Collaborative Historical African Rainfall Model (CHARM) rainfall data set. The data is gridded on an 11 x 11-km basis for the entire African continent and provides daily estimates of rainfall for the period from 1961 to 1996 with a recent update to 2000. LEWS acquired the CHARM rainfall data set, linked it with the generated temperature data, and now uses it for computing historical forage response and deviation of forage from normal. The CHARM rainfall data were provided to RSG in a daily smoothed format from 10-day accumulated historical data, requiring development of a method to create a rainfall "event corrected" dataset from CHARM to allow the data to behave in a more hydrological correct manner in the model.

Using geostatistics to map forage conditions

Link to Figure 4, co-kriging schematic (~34K)

(Back to top)
Because the African data are sampled across large landscapes, we had to develop a technique to estimate regional effects of weather where we did not sample. We chose to use the Normalized Difference Vegetation Index (NDVI) data that provides a measure of green biomass on the ground as seen from the AVHHR satellites. LEWS' automation technology acquires these data for Africa every 10 days from the Earth Resource Observation System (EROS) Data Center where it is processed as part of the Famine Early Warning System Network (FEWSNET). Using the geostatistical technique of co-kriging, we were able to translate greenness data associated with the point-based PHYGROW output into maps of forage standing crop (Angerer et al. 2001) and forage deviation from normal. Co-kriging allows forage data generated for a small set of samples in a large landscape to be coupled with the more spatially rich NDVI data set to interpolate forage responses across a region. For co-kriging to work effectively, a linear relationship must exist between the model forage values and corresponding NDVI data. Since the correspondence between model output and NDVI in co-kriging is spatially dependent, areas with little or no correspondence can be identified, thus allowing the LEWS teams to determine where new sampling points are needed. Currently, the co-kriging is conducted using commercial software (GS+); however, LEWS is designing an automated mapping system based on GSTAT algorithms that will be deployed in the near future. All spatial data generated by LEWS is provided on the Internet and served to the public via the LEWS MAPSERVER web site.

Forecasting 90-day forage conditions

Link to Fig. 5, forage data (~44K)

(Back to top)
The final challenge to meet information needs for decision makers is forecasting. Participants in LEWS have insisted that without a reasonably believable near-term forecast, they would find the technology significantly less valuable as a risk management tool. Therefore, our team explored various forecasting algorithms and found that the AutoRegressive Integrated Moving-Average (ARIMA) model in the SAS software (SAS 1999) would provide a 90-day estimate of forage conditions within normal vegetation sampling error (Kaitho et al. 2003). The ARIMA approach uses modeled forage for past dates and matching historical NDVI conditions along with current forage estimates to predict future grazed standing crop. The values are based on 10-day moving averages and are updated every 10 days. The users of the LEWS system then receive a continuous stream of data that provides an indication of current forage conditions, forage conditions on the same date last year and the forecasted outcomes in the next 90 days. Using correlations with historical scenes of NDVI and working with Columbia University's International Research Institute for Climate Prediction, we believe we will be able to generate 90-day forecasts of NDVI that can provide even richer forecast maps of future forage conditions using the co-kriging technique described earlier.

The LEWS technology in Texas is being tested with the new MODIS multiple-scale NDVI imagery to determine the suitability of the 250-m, 500-m and 1000-m data for use in early warning systems via the co-kriging technology developed in Africa. Early work in South Texas has indicated that the 1000-m NDVI data provides reliable forecast values for a variety of forage types when used in the ARIMA model (Al Hamad 2002). We are currently testing the use of the MODIS data in the desert grasslands and mountains of the Trans-Pecos region of West Texas.

Linking LEWS with decision makers

(Back to top)
As indicated earlier, effective information outreach is critical to the success of the LEWS program. To insure this, special attention must be given to packaging information appropriately for different information users. Currently, LEWS places all analyses for Africa on the African Livestock Early Warning System web site with 10-day updates. In addition, subsets of the analyses are automatically reconfigured for broadcast via WorldSpace satellite radios using African Learning Channel bandwidth and containers from the Arid Lands Information Network (ALIN) and RANET. The maps and minimal narrative go to ALIN for distribution in their network of 100+ satellite radios across East Africa. The full situation reports are distributed by RANET in their larger broadcast container to many more communication nodes. Many of these communication nodes are located in NGO, district and communication offices of early warning agencies of the African LEWS countries. The LEWS reports are also emailed throughout the African LEWS region by the UN Office for Coordination of Human Assistance. These reports are normally condensed versions of the situation reports created by country-level information officers. A consortium of FEWSNET, US Geological Survey, World Food Program, Drought Monitoring Center, Regional Center for Mapping Resource Development and LEWS also produces a monthly Greater Horn of Africa Food Security Bulletin. LEWS provides the pastoral outlook report for that bulletin.

In the Texas LEWS system, our primary communication outlet is the Internet coupled with weekly releases for local newspapers. The outputs from the system are currently being organized by logical geographic areas and designed so that local daily or weekly newspapers can easily download and publish the maps and text. The approach is not unlike that used for The National Drought Mitigation Center's "Drought Monitor," issued over the Internet (Svoboda et al. 2002). Regional training is also provided for State Extension personnel, US Department of Agriculture personnel in the Natural Resource Conservation Service, and interested ranchers. Training sessions are currently limited to 1-2 hours and focus on accessing the Internet, interpreting information, and using the online tutorials. RSG has also acquired experience in online training systems through its development of the web-based Livestock Nutritional Advisory System, designed to complement the information provided by Texas LEWS. This system allows livestock producers to send in feces samples from their animals and enter herd data. Once a sample is received by our laboratory, it is scanned using near-infrared reflectance spectroscopy (Lyons and Stuth 1992) and livestock diet protein and digestibility estimates are made. These are entered into the system, and the producers subsequently receive timely, high quality advisories on performance status of their animals and least-cost feeding recommendations. In the USA, we see this approach as the model for expanded training and increased adoption rates of the LEWS technology. In Africa, we are relying more on distribution of CD training guides that can be played on a laptop in a remote region for one-on-one training in interpretation and dissemination of results. Working in the initial stages with decision-makers and information disseminators alike has allowed the LEWS team to customize both the delivery method and the content of the message to the intended audience.

Passing the "Believability Test"

(Back to top)
The major challenge facing livestock producers' adoption of the information generated from the technology is passing the "believability" test. Two stages to this process must be addressed. First, both scientists and outreach personnel must assure themselves that they are providing consistent, biologically sound information. Extensive verification studies, conducted over several years by our collaborating institutions, built their confidence in the system. Vegetation sampling on 81 of the 300 sites in East Africa indicated a test of correspondence resulting in standard error of prediction of 161 kg/ha, accounting for 96% of the variation in the data set (Jama et al. 2003). Studies on three major plant communities in South Texas over 22 years, indicated that peak standing crop was estimated with an error of 75 kg/ha, accounting for 93% of the variation (Schumann et al. 2002).

Second, the information consumers must also be assured that the information they are receiving is consistent, sound, and useful. The LEWS teams in East Africa have just completed a drought perception survey of pastoral communities throughout the region. Pastoralists were asked to recall their actions in the past 12 months in terms of livestock movement and sales activities. In those sites where water supply was not limiting and conflict did not constrain livestock movement decisions, the LEWS technology is providing well over 90 days of early warning before these pastoral communities perceive that forage conditions are warranting decisions on movement. In some cases, warnings were indicating needed actions well over six months before the pastoralists took action. Discussions with the pastoral communities, after showing the products of the LEWS technology relative to their decision making process, indicated that they value the information but that it needs to be packaged for their particular decision-making environment. Village speakers interpret the LEWS data for village leaders who in turn provide guidance to the community. Posting of graphs and reports at community gathering points or communication nodes such as markets, clinics, livestock supply stores, or NGO/District offices were also identified as important communication methods. In areas with low literacy rates, we are exploring use of rural radio broadcasts of the reports.

Studies in the USA by our research team have indicated that ranchers will use the Internet to access information and analysis they feel is reliable and meets the need of critical decision, particularly when uncertainty prevails (Zander and Stuth 2003). However, they too insist on the "believability" test. Once they acquire the LEWS reports, the information provided must track their own observations. Like Africans, US ranchers want the information in different forms. Small livestock producers seem to want more general data as an "index" of how forage conditions are developing and potentially impacting their operations. Essentially, they want to know "how bad is it?" Larger livestock producers in the Texas LEWS network have requested customized and protected access to the information system where they can specify the longitude/latitude of a rain gauge, link that gauge to one of the PHYGROW representative plant communities, record rainfall events as they occur and then let the output reflect the forage conditions, deviations, percentile ranks and forecasts relative to the variation recorded on their property. Although we have not implemented this level of resolution in the Texas LEWS program, we are currently developing a mechanism for a customized system that has protected access to help livestock producers represent more site-specific conditions than is allowed with the current 25 x 25 km gridded system.

Communicating warnings

(Back to top)
When developing maps of forage conditions, both African pastoralists and Texas ranchers have indicated that maps of classes of conditions would be desirable as long as they provide legible landmarks. Interviews with focus groups have indicated that it is difficult for livestock producers to see changes from "normal" on the ground until there is a plus or minus 20% deviation from long-term normal forage conditions (Jochec et al. 2001). Therefore, we have developed the following mapping categories expressed as the percent deviation from the long-term average daily grazed forage standing crop:

ABOVE NORMAL (>20%)
NORMAL (0 to 20%)
WATCH (0 to -20%)
ALERT (-20 to -40%)
WARNING (-40 to -60%)
EMERGENCY (-60 to -80%)
DISASTER (-80% to -100%)

Link to Fig. 6, forage condition map (~53K)

The delineations appear to be coarse enough to allow rapid appraisal of the maps by non-technical people, yet allow the maps to visually represent forage response over the 7 to 10 day reporting sequence when major changes in forage condition occur at the regional scale (Examples of time sequenced maps can be seen on the African LEWS web site).

Summary and conclusions

(Back to top)
The LEWS technology package has demonstrated that satellite weather and NDVI data can be incorporated into an analytical system to simulate forage conditions over large regions, map those conditions, and forecast likely responses. These value-added data are useful to ranchers, relief organizations, development NGOs, humanitarian groups, extension personnel and pastoral community organizations. The technology is capable of translating complex biophysical conditions into timely information that can be transferred through a diverse communication network ranging from remote NGOs working with solar or locally generated electricity to advanced ranch enterprises with 24/7 Internet capacity.

The LEWS experience to date indicates that more timely and accurate information leads to improved decision-making. Quantification of data that has demonstrated consistency in outcomes provides decision-makers with an objective information base for formulating decisions that affect the lives and livelihood of pastoral communities. This allows a more rational approach to risk management. The paradigm we are using places automated information technology in a stable computing environment with low maintenance needs and the ability to generate output in varied formats relevant to specific audiences. It provides a cost-effective mechanism for using the power of advanced technology while also maximizing the value of high-impact outreach activities through existing information infrastructures. Future advances in space-based monitoring technology coupled with new analytical techniques and greater computing power have the potential to vastly improve upon the LEWS technology package. We believe the LEWS systems can be integrated with other information systems, such as those capturing marketing and biosecurity data, in order to further meet the information needs of livestock producers worldwide.

Acknowledgments

(Back to top)
The Livestock Early Warning System (LEWS) Project is supported by the Global Livestock Collaborative Research Support Program (GL-CRSP) funded in part by the U.S. Agency for International Development (USAID) under grant number: PCE-G-00-98-00036-00. The opinions expressed do not necessarily reflect the views of USAID. Partial funding for this program is also from USDA-NRCS Texas State Office, Noble Foundation, Association for Strengthening Agricultural Research in East and Central Africa - Crisis Mitigation Office, Kenya Agricultural Research Institute, Ethiopian Agricultural Research Organization, Uganda National Agricultural Research Organization, and the Tanzanian Ministry of Livestock and Water Development.

References

(Back to top)
Al Hamad, M.N. 2002. Integration of point biophysical modeling and NDVI data to improve forecasting of near term forage conditions in Texas. Ph.D. diss., Texas A&M University, College Station.

Angerer, J.P., J.W. Stuth, F.P. Wandera, and R.J. Kaitho. 2001. Use of satellite-derived data to improve biophysical model output: An example from Southern Kenya. Paper presented at Sustainable Agriculture and Natural Resource Management (SANREM) Research Synthesis Conference, Athens, Georgia, November 28-30. Online: http://www.sanrem.uga.edu/sanrem/conferences/nov2801/waf/satelitteDataImprovedOutput.htm (accessed 30 June 2003).

Box, G.P., G.M. Jenkins, and G.C. Reinsel. 1994. Time series analysis: forecasting and control. 3rd ed. Upper Saddle River, N.J.: Prentice Hall.

Corbett, J., J.W. Stuth, P.T. Dyke, and A. Jama. 1998. New tools for the characterization of agricultural (crop and livestock) environments: The identification of pastoral ecosystems as a preliminary structure for use in sample site identification. In Proceedings on pastoral early warning systems for Ethiopia, 31-40. Addis Ababa, Ethiopia.

Funk, C., J. Michaelsen, J. Verdin, G. Artan, G. Husak, G. Senay, H. Gadain, and T. Magadazire. 2003. The collaborative historical African rainfall model: Description and evaluation. International Journal of Climatology 23:47-66.

Herman, A., V.B. Kumar, P.A. Arkin, and J.V. Kousky. 1997. Objectively determined 10-day African rainfall estimates created for famine early warning systems. International Journal of Remote Sensing 18:2147-2159.

Higgins, R.W., W. Shi, and E. Yarosh, 2000: Improved United States precipitation quality control system and analysis. NCEP/Climate Prediction Center Atlas Number 7, 40 pp. Online: http://www.cpc.ncep.noaa.gov/research_papers/ncep_cpc_atlas/7/index.html (accessed 25 June 2003).

Hutchinson, M.F. 1991. The application of the thin plate smoothing splines to continent-wide data assimilation. In Data assimilation systems, ed. J.D. Jasper 104-113. Bureau of Meteorology Research Centre (BMRC) Research Report No. 27. Melbourne: Bureau of Meteorology.

Jama, A., M. Kingamkono, W. Mnene, J. Ndungu, A. Mwilawa, J. Sawe, S. Byenkya, E. Muthiani, E. Goromela, R. Kaitho, J. Stuth, and J. Angerer. 2003. Field verification of simulated grazed forage standing crop using the PHYGROW model and satellite-based weather data. USAID Global Livestock CRSP, Research Brief 03-03-LEWS.

Jochec, K.G., J.W. Mjelde, A.C. Lee, and J.R. Conner. 2001. Use of seasonal climate forecasts in rangeland-based livestock operations in West Texas. Journal of Applied Meteorology 40(9): 1629-1639.

Kaitho, R., J. Stuth, J. Angerer, A. Jama, W. Mnene, M. Kingamkono, J. Ndungu, A. Mwilawa, J. Sawe, S. Byenkya, E. Muthiani, and E. Goromela. 2003. Forecasting near-term forage conditions for early warning systems in pastoral regions of East Africa. USAID Global Livestock CRSP, Research Brief 03-02-LEWS.

Lyons, R.K. and J.W. Stuth. 1992. Fecal NIRS equations for predicting diet quality of free ranging cattle. Journal for Range Management 45:238-243.

Nicks, A.D., C.W. Richardson, and J.R. Williams. 1990. Evaluation of the EPIC model weather generator. In EPIC-Erosion/Productivity Impact Calculator, 1. Model documentation, eds. A. N. Sharpley and J. R. Williams, 105-124. USDA-ARS Technical Bulletin 1768.

SAS Institute, Inc. 1999. Procedures guide for personal computers, Version 8 edition. Cary, N.C.: The Institute.

Schumann, K.D., J.R. Conner, J.W. Richardson, J.W. Stuth, W.T. Hamilton, and D.L. Drawe. 2002. The use of biophysical and expected payoff probability simulation modeling in the economic assessment of brush management alternatives. Journal of Agricultural and Applied Economics 33:539-549.

Svoboda, M., D. LeComte, M. Hayes, R. Heim, K. Gleason, J. Angel, B. Rippey, R. Tinker, M. Palecki, D. Stooksbury, D. Miskus, and S. Stephens. 2002. The Drought Monitor. Bulletin of the American Meteorological Society 83:1181-1190.

Stuth, J. W., J. Angerer, R. Kaitho, A. Jama, and R. Marambii. 2003. Livestock Early Warning System for Africa rangelands. In Agricultural drought monitoring strategies in the world, ed. V. Boken. Forthcoming.

Zander, K. and J. Stuth. 2003. Potential adoption rates by producers of an online decision support system for grazing management. National Grazing Lands Conservation Conference, Nashville, Tennessee. Forthcoming.

Xie, P. and P.A Arkin. 1998. Global monthly precipitation estimates from satellite-observed outgoing long wave radiation. Journal of Climate 11:137-164.