Print Masthead

content | navigation | back


United States Department of Agriculture
National Agricultural Statistics Service

images of U.S. agricuture
Search NASS Advanced SearchSearch TipsBrowse NASS by Subject Crops and Plants Demographics Economics Environmental Livestock and Animals Charts and Maps Research and Science Education and Outreach Statistics by State #
You are here:  Home /

  Future Vision for the Use of Remotely Sensed Data


This paper will present a vision for the potential future uses of remotely sensed data, and GIS in NASS. There are several fairly recent new satellite-based sensor systems now available that are promising for NASS applications. NASS is in varying stages of evaluating these new sensors as well as expanding the county level crop acreage estimation and Cropland Data Layer program to ten states for 2003, appropriated funding permitting. NASS benefits from these applications in two ways. The first is an additional input for official county crop acreage estimates. The second is to provide a broad spectrum of GIS proficient data users in the crop production, environmental monitoring and value added remote sensing/GIS industries. NASS also plans to continue crop monitoring on a national/regional scale with AVHRR and MODIS. In addition, this paper will document current uses of GIS in NASS. Among those are geographic display of the Census of Agriculture data, the Cropland Data Layer, the AVHRR based bi-weekly Vegetative Index data, NASS Area Sampling Frames, and aggregated census or survey based data, when appropriate.

In quick summary, the Cropland Data Layer Program is expanding to the 10-20 state level over the next several years, funding permitting of course, and the use of GIS for various applications is continuing to be developed, although at different rates of progress depending upon the priority of the application and staff training in GIS.


Satellite-based sensors having substantial potential for NASS applications that have recently been successfully launched and deployed are: Landsat 7 (U.S. government), MODIS (US government) and IKONOS and QUICKBIRD (private vendors and radar such as AMSR (Japanese). They are listed in priority order for NASS applications. NASS was one of the first users of Landsat 7 in a large scale application with the 1999 crop acreage estimation and mapping program in five states (Arkansas, Illinois, Mississippi, New Mexico and North Dakota). The Landsat 7 data is very similar to previous Landsats and was useable immediately with little change required in the data processing system that NASS uses for crop acreage estimation and mapping projects. Landsat data is virtually ideal for this NASS application with one exception. The temporal repeat overpass of the same area is every 16 days, which is marginal for crop type discrimination, since with cloud cover probabilities, there is usually only one or sometimes no passes in the peak vegetation period. Thus, combining the use of Landsat 5 and Landsat 7 for 8 day coverage is essential. So, Landsat 7 was a most welcome addition for the NASS acreage application.

NASS benefits from a data sharing policy with USDA's Foreign Agriculture Service (FAS) and Farm Service Agency (FSA). For the last several years, this has enabled us to expand the number of states in the program. In the year 2002, we will be at nearly the peak that was achieved in the Agristars program in the early 1980's. As long as 8 day or better coverage is available from Landsat-like sensors, the Cropland Data Layer program will likely grow to around 20 States total, focusing on the Midwest, Delta and Plains.

MODIS is a new satellite sensor that is very similar to the current AVHRR sensor on the NOAA polar orbiting weather satellite system. NASS uses AVHRR to map vegetative index (NDVI) data; this began with the massive Midwest flood in 1993. MODIS will maintain two day coverage while improving the spatial resolution from 1 kilometer (250 acres) to 14.5 acres or 57.5 acres depending on the bands used. Thus, the vegetative index maps that NASS has been putting on the Web for crop monitoring users can be considerably more detailed with MODIS. It will take awhile for NASS to build the necessary infrastructure to handle this new data set providing the National Aeronautics and Space Administration (NASA) and the EROS Data Center can provide real-time biweekly composites. In addition, MODIS is the most likely remotely sensed data set to contribute to a crop yield forecasting system as one input variable (that would still need to be supplemented by ground data, weather data, soil data, crop and soil water balance models etc . . . ). NASS is conducting research with the USDA Agricultural Research Service's Remote Sensing and Modeling Laboratory. This research has begun on a small scale in Illinois (one county) to evaluate MODIS as an input variable for crop yield modeling at the county and sub county level.

IKONOS and QUICKBIRD will be the most questionable for NASS applications of the sensors discussed here. That is not because of data quality which is indeed impressive. IKONOS and QUICKBIRD are is the first very high resolution space-borne systems in the public or private sector outside of any military type systems such as the now publicly revealed CORONA systems. IKONOS has 1 meter resolution for the panchromatic band and 4 meter resolution for
4 multi-spectral bands, similar, to some Landsat bands. QUICKBIRD is similar in sensor parameters with 0.62 meter panchromatic resolution. Thus, at the individual crop field level, the amount and the value of the information about crop condition and identifying area of stress within the field from very high resolution sensors is most impressive. Thus, if the cost (both raw data and information extraction such as vegetative indices) can be made reasonable, individual farmers can use this type of data for monitoring crop stress in fields.

However, large area inventories such as those done by NASS are a different story. The data remains impressive, but the obstacles for successful inventory use remain very substantial. Among these obstacles are: sheer data volumes getting priority in a pointable satellite system, obtaining wall-to-wall coverage for large area projects, effects of non-nadir viewing angles, vendor costs (which are high per unit of land area), and vendors do not typically deliver or set prices for user defined samples. The data volume immediately implies that a sampling scheme is necessary because wall-to-wall inventories of large areas such as states are not feasible without sampling. Thus, a cost effective sampling scheme needs to be developed as a first step before an evaluation could even take place. The vendor would need to be able to price the data for a given sampling scheme so that the cost per area unit of land covered is affordable to the potential user. The sampling scheme would likely go along the satellite scene path lines, which in the case of IKONOS is an 11 km swath, along the straight nadir view. Off nadir views are another issue as they can increase probability of coverage but cause data rectification issues to surface.
One other type of data to mention is satellite based radar data. Over the next decade ARS and NASA are working on soil moisture profiles from space based radar. At the moment, NASS can only follow those research developments. The Japanese developed AMSR sensor on NASA's ACQUA system is one of the most promising with 2 day 50 km global coverage.


First, a brief summary of the current GIS applications in NASS would be helpful. The current GIS applications include area sampling frame construction and maintenance; county estimate theme maps on the NASS Web site; crop season visualization graphics on the NASS INTRANET site; Census of Agriculture graphics on colored paper, the NASS Web site, and on CD's; distribution of crop specific categorization of Landsat data on CD's with metadata; vegetative index maps on a biweekly basis on NASS Web site; Minnesota environmental data on animal waste by six digit watersheds; North Carolina sampling scheme for a specific watershed for farm chemical usage; North Dakota use of the crop specific categorized NASS GIS data layer in conjunction with other external data layers and a university engineering staff to aid in a plant facility location for an agribusiness; and likely several more at the State office level the authors are not currently aware of.

Thus, where is the future use of GIS in NASS headed? Obviously, the scope of applications will depend on the resource commitment (staff available and trained in GIS) and the relative priority of these applications versus other new projects that may be competing for resources at the time. Other than area frame and remote sensing data uses which are obvious fits for GIS and are current applications, there are two new areas where GIS may be on considerable use and benefit to NASS staff or the public at large data user community. The first of these is to provide Census of Agriculture and survey (when appropriate) results to data users by their polygons of interest such as watersheds, townships, planned plant facility circles for evaluation, etc. The issues are: can the data be statistically reliable at those levels, and can it be accurately geo-coded and still protect individual data from disclosure? If these obstacles can be overcome, as the Minnesota 6 digit watershed Census data on animal waste illustrates, then very useful new data products and GIS data layers can be provided to the public by NASS.

The second major area is to use geo-coding of data to aid in designing, conducting and analyzing survey and Census of Agriculture data by NASS statisticians. If geo-coded data from an area sample or a list sample were routinely available and accurate for the given application, statisticians could design better more cost efficient surveys and estimators. The use of location of data could aid in sample design and allocation, data editing and imputation and in estimation (regional and small area such as a county). NASS has examined using list address based geo-coding software. The results as expected, were mixed depending on two major factors. The first factor is what percentage of the list had street addresses versus rural route numbers, etc. The most promising development here has been the addition of 911 emergency address locations for many rural counties, although there is a time lag before the commercial geo-coding databases is updated. The second factor is average farm size in a State. Using one location for a large farm with multiple and non-contiguous land parcels leaves some accuracy issues depending on the particular GIS application. Thus western States will pose a bigger challenge than northeastern States. When NASS publishes geo-coded results, such as for Census of Agriculture (political districts, 5 digit zip codes, etc.) metadata is provided describing the method of geo-coding. The users can then decide if the accuracy is sufficient for their application.

One of the promising developments is a joint research project between the Farm Service Agency (FSA) and NASS. For a pilot area in Nebraska, the FSA common land unit GIS data base and the NASS Cropland Data Layer and the NASS Area Sampling Frame sample units have all been merged into one GIS file to examine the efficiencies and strengths of the combined data sets. NASS has also begun to work with the Animal and Plant and Health Inspection Service (APHIS) regarding the location of large livestock operations. In all cases, NASS will not disclose any individual operation locations or data and only use them for statistical sampling and aggregation purposes.



Last modified: 05/27/09