National Raster of Subsurface Tile Drainage
Two national 30-m resolution rasters of subsurface tile drainage were created from the following sources:
and geospatial data of:
The differences between the two versions are the sources for land cover and county boundaries:
The goal was to produce a geospatial representation of Zachary Sugg’s (2007) county tabular data of subsurface tile drainage areas. Sugg’s work focused on improving the 1992 National Resources Inventory (NRI) county database of the subsurface drainage areas in the Midwest states (IL, IN, IA, MO, OH). He based his work on the theory that using GIS to define areas with row crops grown over poorly drained soils in known densely tiled regions had the most potential for improving previous subsurface tile drainage estimates (Sugg, 2007). Sugg further improved his Midwest estimates with state-level percentages based primarily on the 1987 USDA report, “Farm Drainage in the United States: History, Status, and Prospects”, and for Ohio, the 1998 Ohio State University Extension Bulletin 871-98 as well. In addition, he consulted with NRCS specialists to acquire local expertise and adjusted his county estimates accordingly (Sugg, 2007). For the non-Midwest states, the county estimates from the USDA/NRCS National Resources Inventory of 1992 (NRI 92) were applied into his data file.
Sugg used the “row crops” class from the National Land Cover Dataset (NLCD) 1992 and the dominant soil drainage classes for the original STATSGO soil map units to define poorly drained soils. The STATSGO soil drainage classes that were used were “somewhat poor”, “poor”, “very poor”, “poor-to-very poor”, and “very poor-to-poor”.
While the initial intent was to create a mapped representation of Sugg’s county tabulations of subsurface drainage representing the late 1980s to early 1990s, there was also interest in using more up-to-date mapped cropland. For this reason, the approach was applied to both the mapped cropland from the NLCD 1992 and 2011 (and subsequently the 1990 and 2010 county boundaries, respectively).
Obtaining the source Sugg used for mapped row crops was straight forward; however acquiring the exact same soils data was problematic. Sugg applied dominant soil drainage classes for soil map units in USDA/NRCS’s State Soil Geographic Database (STATSGO) database (USDA, 1994). This original database is no longer accessible publically as it was replaced by USDA/NRCS’s STATSGO (version) 2 shortly after Sugg’s report was published. Even though the original STATSGO data were found to be internally accessible, limited time and resources made it unfeasible to process. Instead, a national 100-m resolution raster of soil drainage classes from STATSGO 2 (Michael E. Wieczorek, written commun., July 24, 2015) was used to extract the poorly drained soils in developing this dataset.
Out of the 3073 county records in Sugg’s file, 2622 or about 85% is equal to the NRI 92’s estimate for subsurface drainage areas, leaving 451 counties or about 15% that reflects Sugg’s estimates. In the version used to develop this dataset, 11 counties of the 451 were corrected in Illinois (Sharon L. Qi, USGS, written. commun., July 2015; Mark B. David, University of IL Urbana-Champaign, written commun., August 2008).
Brief Description of the Procedure:
All geoprocessing was set at the 30-m resolution.
Mapped cropland areas classified as “row crops” in the 1992 NLCD and “cultivated crops” class in the 2011 NLCD, were overlaid (separately to create 2 different datasets), with the areas characterized as having poorly drained soils (STATSGO2 soil map units having dominant drainage class of “somewhat poorly drained”, “poorly drained”, or “very poorly drained”). For the Midwest counties that reflected Sugg’s (and UIUC’s) subsurface drainage area estimates, these intersecting pixels were then identified as area onto which Sugg’s county values would be apportioned. For the counties that reflected the raw NRI 92 county estimates, federal lands were also overlaid to exclude these areas from the intersecting cropland with poor drainage. This is because NRI surveys excluded federally-owned land.
The output raster from these overlays resulted in a national raster of potentially tile-drained areas – the cropland with poorly drained soils (output raster = “CROP_POORDRN”). A county boundaries raster was then overlaid with the CROP_POORDRN raster to calculate the county areas (in square meters) of cropland with poorly drained soils (fieldname = “CROP_POORDRN_M2”). The resulting table was combined with Sugg’s county table of subsurface tile drainage numbers (“SUGG_M2”) so that the ratio of SUGG_M2 to CROP_POORDRN_M2 could be calculated. This county ratio field was joined to the county raster attribute table, and a new raster was created from the ratio using the LOOKUP tool. Because the raster is at the 30-m resolution, each pixel has an area of 900m2, so the ratio raster was multiplied by 0.0009 to convert the unitless raster to square kilometers. The resulting raster represents the final raster of subsurface drainage areas, with each pixel expressed in square kilometers.
Although an attempt was made to replicate Sugg’s work of combining the mapped row crops with poorly drained soils, the source for poorly drained soils used for this dataset was not what Sugg used. The replacement of STATSGO2 for the original STATSGO introduced some inconsistencies because of changes made to soil map units and soil parameters between the two versions.
In addition, in regard to the county subsurface drainage estimates, Sugg (2007) states that the revised NRI 1992 data (in the Midwest) are “not intended to be final” because the “best guess” estimates need validation. Also, some of the derived numbers are based on older publications (at the state level) and surveys.
Furthermore, the methodology applied in creating this dataset introduced additional limitations. In the perfect world, the county ratios would be 1.0, in other words, Sugg’s estimated county area of subsurface tile drains would be equal to the county GIS-derived area of cropland grown over poorly drained soils. Ratios less than 1.0 would equate to less than 900 meters of the 30m-by-30m pixel is tile drained, which doesn’t reflect reality. Moreover, ratios greater than 1.0, produce a different type of error. For example, when a watershed boundary is overlain with a number of subsurface drain raster pixels that have values greater than 900 (the true area of the pixel, and that had ratio pixels greater than 1.0 in the same geographic 30mx30m area), the total sum of the intersecting subsurface drain pixel values could exceed the area of cropland in the watershed. This situation occurred in a few watersheds in the Midwest, and were revised to equal the area of the cropland in the watershed.
Because this dataset (version using 2011 NLCD row crops and counties) was used to characterize stream sites in the Midwest for a nitrate model for which the results will be published, the plan is to publish the 2011 rowcrops-Version 2 subsurface tile drainage dataset (national raster). However, the intent with the NAWQA gis group is to ultimately create a refined version using higher detailed soils data (Soil Survey Geographic Database, or SSURGO) and possibly integrate crop-specific information from the National Agricultural Statistics Services’ Cropland Data Layer (Michael E. Wieczorek, oral commun, July 2015).
Naomi Nakagaki (email@example.com)
Citations and their contributions if used to develop the dataset:
Gronberg, J.M., 2012, 30-meter-resolution grid of 1990 counties, conterminous United States, raster digital data, accessed December 2013, at http://water.usgs.gov/lookup/getspatial?sir2012-5207_co1990g.
Contribution: 1990 county boundaries were used to apportion the county areas of subsurface drainage areas
Gronberg, J.M., (unpublished), 30-meter-resolution grid of 2010 counties, conterminous United States, raster digital data, acquired April 5, 2012
Contribution: 2010 county boundaries were used to apportion the county areas of subsurface drainage areas
Homer, C.G., Dewitz, J.A., Yang, L., Jin, S., Danielson, P., Xian, G., Coulston, J., Herold, N.D., Wickham, J.D., and Megown, K., 2015, Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogrammetric Engineering and Remote Sensing, v. 81, no. 5, p. 345-354
National Atlas of the United States, 2005, Federal Lands of the
United States, vector digital data, accessed January 2015,
Contribution: mapped Federal Lands to exclude these areas within counties where NRI subsurface drainage areas were apportioned
Sugg, Zachary, 2007, Assessing U.S. Farm Drainage: Can GIS Lead to Better
Estimates of Subsurface Drainage Extent?, World Resources Institute,
Washington D.C., at http://pdf.wri.org/assessing_farm_drainage.pdf
Contribution: county areas of subsurface drainage areas, from attributes “Best Guess (ac)” and “NRI 1992 (ac)”, and “Source”.
U.S. Department of Agriculture, 1987, Farm Drainage in the United States: History, Status, and Prospects. USDA-ERS Miscellaneous Publication Number 1455. Washington, D.C.
U.S. Department of Agriculture, 1994, State Soil Geographic (STATSGO) Data Base, Data use information. U.S. Department of Agriculture Miscellaneous Publication Number 1492, Fort Worth, Texas, 35 p.
U.S. Department of Agriculture, 2007, U.S. general soil map (STATSGO2) data: Natural Resources Conservation Service geospatial data, accessed April 2007.
Contribution: the mapped soil units and soil drainage classes were used to develop the soil rasters developed by Michael E. Wieczorek (see below).
U.S. Geological Survey, 2007, National land cover dataset 1992 (NLCD1992), raster digital data, accessed July 14, 2015, at http://www.mrlc.gov/nlcd92_data.php
Contribution: mapped “row crops” classification
U.S. Geological Survey, 2014, National land cover database 2011 (NLCD 2011), raster digital data, accessed February 15, 2015, at http://www.mrlc.gov/nlcd2011.php
Contribution: mapped “row crops” classification
Wieczorek, M.E., (unpublished), 100-m resolution raster of dominant soil drainage classes from STATSGO 2, acquired 7/24/15.
Contribution: this raster defined the poorly drained areas in the final tile drains raster; pulled mapped soil map units represented mostly by 1) somewhat poorly drained (value = 5); 2) poorly drained (value = 3), or very poorly drained (value = 6)