Frequently Asked Questions

Q: Are there any peer-reviewed, scientific publications available that describe the science underlying Grass-Cast?

A: Yes, here are citations and links for several key papers written by members of the Grass-Cast team.

Q: Are the data used to develop Grass-Cast available to the public?

A: Most of the data underlying Grass-Cast are available, thanks to cooperators’ who gave us permission to share some of their data publicly. Datasets for which we have permissions to share are available through USDA’s Ag Data Commons: https://data.nal.usda.gov/dataset/grass-cast-database-data-aboveground-net-primary-productivity-anpp-climate-data-ndvi-and-cattle-weight-gain-western-us-rangelands/resource/87d899d5-c103-4485-a8db-068327805ef6. For any remaining data needs or requests, please contact the Grass-Cast team coordinator, Dr. Dannele Peck, at dannele.peck@usda.gov.

Q: Can you zoom in on the Grass-Cast maps to look closer at the counties?

A: We now offer two options for viewing the Grass-Cast maps: 1) as a static image, or 2) as a zoomable map.

1) If your internet connection is slow or you are on your cell phone, we recommend viewing the static image. On your cell phone, you can zoom in and out from the static image using your fingertips, just like you would with a photo. Or on your computer, you can manually zoom in and out from the static image by holding the Ctrl button and tapping the + or - key on your keyboard./p>

2) If your internet connection is fast enough, try our new (still in progress) zoomable version of Grass-Cast. Start by clicking on the search icon , entering your location (town, zip code, cross-street) in the search box, and clicking on the most relevant location in the drop-down list. Then you can use the + and - buttons in the upper-left corner of the left-most map to zoom in or out. You can zoom in as close as an individual 6.2-mi2 (10-km2) grid cell. However, we recommend zooming out to a broader level (e.g., county-level) to see what colors are most common in the broader region.

Q: Do you average the precipitation data across as many weather stations as you have access to per grid cell?

A: Currently, the color assigned to a grid cell is estimated for a single location within that cell, particularly the centroid or geographic center. The Grass-Cast model uses the latitude/longitude for that centroid to pull precipitation data from the closest weather stations to it, and takes a spatially-weighted average of them.

Q: For grid cells in white or no color, will there ever be sufficient data to provide a Grass-Cast outlook, or do you foresee those locations being left out of Grass-Cast for some time?

A: Many of the grid cells in white (or no color at all) are excluded from Grass-Cast because they are dominated by a vegetation type that interferes with the statistical signal between precipitation and the greenness index (NDVI).Examples include cropland (dryland or irrigated), shrublands, and forests. For counties with these vegetation types, we do not foresee being able to produce Grass-Cast in the near future. For other counties in white (or no color), the remotely-sensed AVHRR NDVI dataset (1982-2015) has too many gaps in it. For these, as enough time passes, we may be able to build up a long enough dataset to someday provide a Grass-Cast outlook. Alternatively, we have started looking at MODIS NDVI data, which has a shorter record (2000-2017) compared to AVHRR, but has data for some counties where AVHRR has none. We are in the process of comparing MODIS NDVI to AVHRR NDVI to see if it can be useful.

Q: What qualifies as "above" or "below" normal precipitation?

A: For a given grid cell, the Grass-Cast team ranks its historical precipitation data from lowest to highest, then splits this ranked dataset into three equal-portion subsets (terciles): the upper one-third, middle one-third, bottom one-third of the dataset. Thus, if the upcoming growing season is forecasted to have "above-normal" precipitation, this means it is likely to resemble or fall within the top 1/3rd (upper tercile) of the historical dataset. If the upcoming growing season is forecasted to have "near-normal" precipitation, it is likely to resemble the middle 1/3rd (middle tercile) of the historical dataset. And if the upcoming growing season is forecasted to have "below-normal" precipitation, it is likely to resemble the bottom 1/3rd (bottom tercile) of the historical dataset. (For an illustration of this concept, courtesy of NOAA, see: https://www.ncdc.noaa.gov/monitoring-references/dyk/ranking-definition

Q: The spring of 2018 was long and cold in much of Nebraska and the Dakotas. Does Grass-Cast account for the effect of cold spring temperatures on grassland production, including its impact on cool-season versus warm-season grasses?

A: The Grass-Cast model does account for observed temperatures, including the cool spring of 2018. These cool temperatures influence our prediction of actual evapotranspiration (AET) for the entire growing season. In turn, AET influences our prediction of above-ground net primary productivity. However, Grass-Cast cannot distinguish warm-season grasses (C4) from cool-season grasses (C3), so it cannot forecast how these two groups of plant types will respond differently to cool spring temperatures. Grass-Cast instead forecasts the impact on the entire plant community as a whole.

Q: The "Bomb Cyclone" during the winter of 2019 brought severe flooding and ice-related damage to many parts of Nebraska and the Dakotas. Does Grass-Cast account for the damage this event caused to the soils and plant communities on grazing lands, and how it will affect vegetation growth during the upcoming growing season?

A: The Grass-Cast model will account for the unusual precipitation and temperatures that occurred during the Bomb Cyclone of winter 2019. So Grass-Cast will account for areas that received a lot of precipitation and therefore might have more soil moisture from precipitation. It will also account for areas that experienced colder than normal temperatures, and how this affected evapotranspiration in the region. However, Grass-Cast model will NOT account for flooding of grazing lands caused by ice-jams, breached levies, or rivers overflowing their banks, which caused much of the severe flooding. Grass-Cast will also not capture any physical damage to grazing lands caused by this type of flooding - such as soil erosion, plant die-offs due to lack of oxygen or contaminants in the water. Therefore, Grass-Cast will probably over-estimate how well vegetation grows in areas that were damaged during the 2019 floods. This is a perfect example of why Grass-Cast should be combined with a landowner's intimate knowledge of their landscape, including its unique characteristics and conditions at the local scale and how these differ from those at the larger landscape scale.

Q: Why do neighboring grid cells have different colors within an individual map? What is so different between two neighboring cells, given they are only 10km x 10km (6 mi x 6 mi) in size? Similarly, when neighboring grid cells do have the same color, they sometimes form an interesting shape on the map, suggesting they share something in common, such as soil type. But then this shape does not stay the same across the 3 maps, suggesting they don't have this in common. Why does this happen?

A: Soil depth and texture may in fact differ between adjacent grid cells. In such cases, your grid cell may simply be more sensitive to precipitation than a neighboring grid cell. Also, historical weather might differ between adjacent cells. In fact, it is rare for two grid cells to have the exact same weather through time. For these two reasons, neighboring cells may have different colors within the same map (i.e., may have a different forecasted changes in their pounds-per-acre). Differences in historical weather are also important because we use your grid cell's unique historical data to construct its future weather over the rest of the growing season. For example, to construct your grid cell's "below-average precipitation" scenario (the far-right map), we first identify the 12 driest years in your grid cell's history. Then we use the daily weather from these 12 driest years to construct your grid cell's future weather for the "below-average precipitation" scenario. After simulating these 12 different future weather possibilities, we take the average of their results to determine which color to assign your cell. When we use this procedure on a neighboring cell, its 12 driest years in history might be different than your cell's 12 driest years. For example, 2011 may have been one of the 12 driest years in your grid cell, but not in a neighboring grid cell (perhaps 2011 was only the 13th or 14th driest year there). In this case, we would use the daily weather from 2011 to help construct your grid cell's future weather (for the "below-average" scenario), but would not use it to construct the neighboring cell's future weather. These differences in future weather may result in your grid cell falling within a different color-category than a neighboring cell, at least for the "below-average precipitation" scenario. For the "above-average precipitation" scenario (the far-left map), it's possible that your cell and a neighboring cell have the same 12 wettest years in history. In this case, if the daily weather observed during these 12 wettest years is similar enough in your cell and the neighboring cell, then they might fall within the same color category for the "above-average precipitation" map, even though they did not fall within the same color category for the "below-average precipitation" map.

Q: How is Grass-Cast different than remotely-sensed estimates of vegetation production, such as the NDVI (Normalized Difference Vegetation Index)?

A: Grass-Cast estimates are not the same as remotely-sensed estimates of grassland production, such as the NDVI. Grass-Cast does use NDVI, but only as one of several inputs to a four-step modeling process (see the Science Webinar Recordings to learn more). Grass-Cast therefore adds value to the remotely-sensed NDVI data by incorporating additional information, such as weather, an ecological model of how grasses grow (known as DayCent), and historical vegetation clippings data from approximately a dozen different locations in the Great Plains region.

Q: Does the Grass-Cast map show just the estimated production for grasses, or for forbs and shrub too?

A: Grass-Cast provides an estimate of vegetation production in GENERAL for a given 6x6-mile grid cell. It is not smart enough to separate out how much of that production might be made up of grasses, forbs, or shrubs. Similarly, it cannot tell the difference between palatable and unpalatable species. That said, Grass-Cast's production estimates are most accurate in grid cells that are dominated by grass species. This is because Grass-Cast relies upon the DayCent model, which was designed for ecosystems dominated by grass species. Similarly, Grass-Cast also relies upon NDVI (Normalized Difference Vegetation Index) in some, but not all, of its modeling steps. It is well known that NDVI is more accurate in grassland-dominated ecosystems than in shrubby ecosystems-because the greenness signal from shrubs does not correspond as strongly with how well it is growing. NDVI is also easily confused by evergreen trees and bare ground. Wherever NDVI struggles to accurately detect vegetation growth, Grass-Cast will also struggle to provide accurate estimates of vegetation productivity. That said, Grass-Cast does still work in ecosystems with a mix of grasses, forbs, and shrubs present-just not as accurately as it would in areas with grass species only. And, again, Grass-Cast lumps them all together in its estimate of vegetation production, so it provides a GENERAL estimate of how much more or less productive we expect the site to be during the upcoming growing season, relative to its long-term average.