Landscape Ecology

, Volume 29, Issue 1, pp 81–95 | Cite as

Agricultural expansion: land use shell game in the U.S. Northern Plains

  • Carol A. Johnston
Research Article


Land area planted to row crops has expanded globally with increased demand for food and biofuels. Agricultural expansion in the Dakota Prairie Pothole Region (DPPR), USA affects a variety of agricultural and non-agricultural land-use types, including grasslands and wetlands that provide critical wildlife habitat and other ecosystem services. The purpose of this study was to quantify recent changes in rural land cover/land use, analyze trends, and interpret results in relation to climate, agronomic practice, and ethanol production. The primary data sources were 1980–2012 statewide cropland data from the U.S. Department of Agriculture (USDA) National Agricultural Statistics Service, and the USDA Cropland Data Layer, produced annually for the DPPR from 2006 through 2012. Area planted to corn or soybean row crops increased, and small grain (e.g., wheat, barley) area decreased significantly over the analysis period. Corn and soybean expanded by 27 % in the DPPR between 2010 and 2012 alone, an areal increase (+15,400 km2) larger than the U.S. state of Connecticut. This expansion displaced primarily small grains and grassland (e.g., pastures, haylands, remnant prairies). Grassland regularly exchanged land with corn and soybean, small grains, and wetlands and water. Corn and soybean had high inter-annual self-replacement values (68–80 %), and continuous corn/soy row cropping was the second most common combination over a three-year period, ranking after continuous grassland. Small grain self-replacement values were only 22–35 %, indicating frequent relocation in the landscape. Temporary gains in wetland and grassland area were attributed to unusually wet climatic conditions and late snowfalls that prevented crop planting. Nearly all of the region’s ethanol refineries were located where corn and soybean crops constituted 50 % or more of the land area. Quantification of grassland losses in the U.S. Northern Plains requires evaluation of all land uses that interact with grasslands, and a longer term perspective that incorporates grassland as part of a normal land-use rotation.


Grassland Corn Soybean Small grains Dakota Prairie Pothole Region Biofuels Crop rotation 


Ecologists have recently raised concerns about the loss of grassland and wetland area due to changing agricultural practices in the U.S. Northern Plains. The area planted to corn or soybeans (corn/soy) in North and South Dakota more than tripled between 1980 and 2011 (USDA NASS 2013), increasing from only 5 % of the two-state area in 1980. Cropland encroachment into grasslands and wetlands could have continental-scale implications to wildlife habitat and declining bird populations. The region has a key location on the Mississippi flyway, providing critical habitat for migratory waterfowl (Stephens et al. 2005). Grassland losses impact upland birds, including economically important game species such as pheasants (Cunningham and Johnson 2006; Stephens et al. 2008). Northern Plains grasslands and wetlands provide many economic as well as environmental benefits, such as forage for grazing livestock, hunting opportunities, and carbon sequestration (Tan et al. 2005; U.S. GAO 2007). Native grassland that has never been plowed (i.e., virgin sod) is particularly valued, because plowing of native grassland releases stored carbon and drastically alters its plant community, belowground fauna, and soil properties (Samson et al. 2004; Searchinger et al. 2008).

Furthermore, the environmental impacts of agricultural expansion are broader than just habitat loss. Agricultural intensification reduces soil ecosystem function and decreases soil carbon (West and Post 2002; Culman et al. 2010). Erosion from row crops degrades soil quality and pollutes streams with sediment and phosphorus (Vaché et al. 2002; Riseng et al. 2011). Corn uses more pesticides and nitrogen fertilizer than any other crop extensively produced in the region and is a major contributor to groundwater and river water pollution (National Research Council 2003; Langpap and Wu 2011). Watershed modeling in Iowa and the James River Valley of the DPPR showed increased stream loadings of nitrogen and phosphorus that “could justify serious concerns regarding increased corn rotations” (Secchi et al. 2011; Wu et al. 2012).

Agricultural practices will likely continue to intensify under future demands for food and ethanol (Tilman et al. 2011; Wallander et al. 2011). The 19 active ethanol refineries in the eastern Dakotas have the capacity to produce 5.21 billion liters of ethanol annually, 10 % of 2011 total U.S. ethanol production (RFA 2013), and South Dakota ranked fifth among the United States for ethanol production in 2011 (U.S. Energy Information Administration 2013).

Climate extremes also influence agricultural land uses in the Northern Plains. Prolonged periods of drought, snow cover, or excess moisture influence not only the type of crop planted, but whether a crop can be planted at all. The Northern Plains climate is harsh and variable, which limits the region’s agricultural practices. Climate also influences the abundance of wetlands in the region, many of which are naturally ephemeral (Johnston 2013).

Germane to these topics are two recent studies that have reported extensive conversion of grassland to corn, soybeans, and wheat in the central U.S. (Faber et al. 2012; Wright and Wimberly 2013) based on geospatial data from the U.S. Department of Agriculture (USDA) cropland data layer (CDL). Grasslands are certainly being lost in the central U.S., but grasslands are also part of normal farming practices, commonly incorporated into rotations as a means of letting the soil recover from intensive crop management. Quantification of grassland conversion in the central U.S. requires evaluation of all land uses that interact with grasslands, not just selected croplands, and a longer term perspective that incorporates grassland as part of normal agricultural land-use rotation.

The purpose of this study is to quantify recent changes in rural land cover/land use, analyze trends, and interpret results in relation to climate, ethanol production, and agronomic practice in the Dakota Prairie Pothole Region (DPPR). This region was selected because it contains large tracts of remnant native prairie and wetlands that are highly valued for wildlife habitat and other environmental benefits, but is experiencing rapid agricultural expansion. The region has acceptable growing conditions for non-irrigated corn, yet corn crops constitute a much smaller proportion of the landscape than in the central U.S. Corn Belt, providing the opportunity for expansion. Specific questions addressed are:
  • What are the geographic distribution, duration, and diversity of agricultural and non-agricultural land uses within the region?

  • What are the long-term (1980–2012) and short-term (2006–2012) trends in area planted to major crops?

  • What land-use sequences occur over a two- to three-year time span? Are these transitions changing over time?

  • Where is agricultural expansion occurring? What land uses have been displaced?

  • How has the existence of grassland and other land uses been affected by agricultural expansion?

The main data source for this analysis was the CDL, a digital land use/land cover map produced annually from satellite imagery which shows the type and location of crops planted in the conterminous U.S., along with other land-cover types such as lakes, grasslands, and wetlands, (Boryan et al. 2011). In contrast to previous CDL analyses by Faber et al. (2012) and Wright and Wimberly (2013), which focused on grassland losses, this paper analyzes transitions among all major crops, non-crop vegetation (forest, grassland, wetlands), and water. The paper also examines land-use sequences so as to better distinguish new plowing of perennial land cover (e.g., pasture, prairie, wetlands) from the temporary, often intentional inclusion of herbaceous vegetation in a crop rotation (e.g., fallow, cover crop).

Materials and methods

Study area

The DPPR is bounded by Canada, the Red River of the North, and the Missouri River (Fig. 1). Continental glaciers deposited a nearly level plain of thick till over the DPPR, but the topography is more complex in the Turtle Mountains (Fig. 1: 46b) and Missouri and Prairie coteaus (Fig. 1: 42a, 42d, 46k, 46l), which resulted from collapse of superglacial sediment. The glacial till soils are predominantly fine-textured loams, but sandy soils occur on glacial outwash and glacial lake deltas (Fig. 1: 46d, 46j). Hilly and sandy areas have been considered marginal for the production of row crops, and were historically retained as grassland for pasture or the harvesting of hay.
Fig. 1

Distribution of non-crop vegetation (grassland, forest, wetlands) and water within the DPPR, 2012, computed as the zonal summary of CDL data by level IV ecoregion. Level III ecoregions are northwestern glaciated plains (42), Northern Glaciated Plains (46), Western Corn Belt plains (47), and Lake Agassiz plain (48). Lower case letters denote level IV ecoregions, defined at:

The DPPR is agriculturally productive due to fertile soils and precipitation that is usually adequate for corn growth without irrigation. Average annual maximum temperature ranges from 16.1 °C at the southern edge of the region to 7.8 °C along the Canadian border. Cumulative growing degree days for corn at the region’s northern edge are only 945 °C; growing degree days <1100 °C limit corn cultivation to rapidly maturing varieties (Ransom 2004). Precipitation is greatest (~700 mm/year) in the southeast and least (~350 mm/year) in the northwest portion of the DPPR. Corn requires 460–560 mm of soil moisture during most growing seasons to achieve maximum yield potential without irrigation (NDSU 1997).

North and South Dakota are rural states, with a human population density of only 4 people per km2 (U.S. Census Bureau 2010). There are relatively few urban areas in the DPPR, but the rural road network is well developed, with roads circumscribing most square mile sections.

Data sources and pre-processing

Data sources used included GIS layers (ecoregions, CDL for 2006–2012), tables from the USDA National Agricultural Statistics Service (NASS) summarizing the area planted to major crops in North and South Dakota (1980–2012), an industry listing of ethanol factories, and climate data (Palmer Hydrological Drought Index 2004–2013). The DPPR study area was defined using a GIS ecoregion database for the Dakotas (, subdivided into level III (coarser, denoted by numbers) and level IV ecoregions (finer, denoted by letters).

Long-term crop data, 1980 through 2012, were analyzed for corn, soybeans, and wheat from statewide statistics published by the NASS, which are obtained by farmer interviews and probability surveys (USDA NASS 2013). Locations of ethanol manufacturers in the DPPR were obtained from the Renewable Fuels Association (, and I used the placemark tool in Google Earth to pinpoint factory locations for import into ArcGIS.

Seamless CDL GIS data for 2006–2012 were downloaded from the CropScape portal ( The CDL design and products have been relatively consistent since 2006, the main change being an increase in spatial resolution with the adoption of new satellite platforms: 56 × 56 m pixels through 2009 and 30 × 30 m pixels for 2010 through 2012 (Boryan et al. 2011). Category codes and class names were standardized in the January 2012 re-release of the entire catalog of historical CDLs.

Rural roads were inconsistently mapped by the CDL, and their area was over-represented by the 56-m pixels on the 2006–2009 data products. Therefore, I created a mask consisting of any pixel that had been mapped as developed on any of the seven analysis dates, and excluded that area from transition and short-term trend computations.

Any geospatial data source contains errors, and those errors are compounded when comparing multi-temporal data (Cunningham 2006; Fang et al. 2006; Shao and Wu 2008). Several sources of error could have impacted the CDL analyses. First, the CDL has inherent errors. Metadata for each CDL product in North and South Dakota showed that corn and soybeans were mapped with producers’ and users’ accuracies ≥89 % in 2007 through 2012, but accuracies in 2006 were slightly lower. Second, the CDL classification detail changed over the years for grassland. Lands classified as #171 Grassland Herbaceous in 2008 were subdivided into #171 Grassland Herbaceous and #37 Other Hay/Non-Alfalfa in 2009, and class #181 Pasture/Hay was merged with class #171 Grassland Herbaceous as of 2011. Potential errors due to these inconsistencies were avoided by merging all grassland/herbaceous, pasture, and hay (except alfalfa) into a single grassland class (Table 1). Third, positional error could produce spurious transitions if the location of grid cells was mismatched from year to year. Mismatch errors were minimized by setting the snap environment in ArcGIS to ensure grid cell alignment and by inspecting consecutive CDL data layers to ensure the consistent position of durable features from year to year. Fourth, the change in CDL spatial resolution from 56-m in 2009 to 30-m pixels in 2010 could influence transitions and rotations computed across this time period (2009–2010, 2008–09–10, 2009–10–11). I retained the coarser pixel resolution (56-m) in these comparisons, which may have favored land uses occurring in large, contiguous patches over small, interspersed land uses. Fifth, land covers that were sparse or narrow were prone to remote sensing edge effects, filling a grid cell in one year but not the next. Rural roads were particularly subject to such edge effects because they are typically only a single grid cell wide; this potential error was alleviated by masking out all developed lands as noted above.
Table 1

Land use/land cover in the DPPR extracted from the 2012 CDL layer, and applicable codes from any of the seven CDL layers


2012 Area (km2)

CDL codes

Row crops






1, 12, 13, 241

Small grains



22, 23, 24, 26, 225, 236







 Other small grains


25, 27, 29, 33, 34, 35, 38, 39, 205

Other crops




 Dry beans, lentils, peas


42, 52, 53










 Sugar beets



 Other row and specialty


4, 43, 44, 47, 49, 57, 58, 59, 60, 68, 182, 229, 246




37, 62, 171, 181



87, 190, 195



83, 111

 Forest and shrubland


63, 64, 141, 142, 143, 152

 Fallow/Idle cropland



 Developed and barren


65, 82, 86, 88, 121, 122, 123, 124, 131

For definition of individual codes, see CDL metadata (

The Palmer Hydrological Drought Index, obtained from the National Climatic Data Center, was used to show unusually dry or wet soil conditions (NCDC 2013). Drought is shown by negative numbers (−4 is extreme drought) and excessive wetness is shown by positive numbers (+4 is extremely moist).

Data analysis

Regime shift detection was used to analyze long-term trends (1980–2012) based on the statewide NASS statistics (USDA NASS 2013). Regime shift detection identifies rapid changes from one relatively stable mean value to another using a sequential t test within a data time series (Rodionov 2006). Shifts in means were analyzed using the following parameters: significance level = 0.05, cut-off length = 6 years, Huber’s weight parameter = 2, and red noise (serial correlation) estimation using the inverse proportionality with four corrections (IP4) method.

All geospatial data layers were analyzed in ESRI® ArcGIS™ 10.1 as follows. Downloaded CDL data were clipped with the DPPR study area polygon. Many of the CDL classes mapped were minor in extent, and some were not applied in all years, so the CDL crosswalk document was used to combine equivalent classes having different codes (Table 1). Because corn and soybeans are typically grown in rotation, I combined their land areas for most analyses, keeping them separate only for the 2012 land use/land cover area summary (Table 1), the computation of land-use diversity, and the percentage of lands in major crops relative to proximity to an ethanol plant (see next paragraph). The duration of presence for corn/soy, small grains, and non-cropland vegetation or water was determined by generating binary (0 = absent, 1 = present) data layers for each land use and date, then summing across dates.

To visualize the geographic distribution of major crop groups within the region, I aggregated the 30-m 2012 CDL data by computing the proportion of small grain or corn/soy land within a coarser grid, each grid cell representing a 31.5-km2 area (i.e., 187 × 187 pixels). Land-use diversity was computed by using “variety” as the statistic for non-overlapping 31.5-km2 neighborhood blocks. The proportion of non-cropland vegetation or water within level IV ecoregions was computed as zonal statistics in ArcGIS. Land use proximal to ethanol plants was determined by generating 30 km buffers around the point locations of each ethanol plant, then clipping the 2012 CDL data to summarize land use both within and outside of the buffer circles.

Cumulative areas of major land uses (corn/soy, small grains, other crops, grassland, wetland/water, forest) were extracted from the seven CDL layers (2006–2012), and their trends were analyzed using linear regressions and the Mann–Kendall test for monotonic trend in a time series using R statistical software (version 3.0.0) (R Core Team 2013). After masking out developed lands, transition (two-year) and rotation (three-year) rates were computed among the major land-use groups for each pair or triple of successive CDL dates using the ArcGIS raster calculator.

To map the gain of corn/soy land between 2011 and 2012, I subtracted the 2011 corn/soy binary layer from the 2012 corn/soy layer, resulting in three possible pixel values: −1 = corn/soy in 2011 but not in 2012, +1 = corn/soy in 2012 but not in 2011, 0 = no change (i.e., corn/soy in both years OR not-corn/soy in both years). For visualization, I then aggregated the data into a coarser block by taking the mean pixel value within a 31.5 km2 area.

To determine locations which were newly planted to corn/soy for the first time in 2011, I summed the binary corn/soy files for 2006–2010, retaining only those pixels that had not been planted to corn in any of the five dates. I intersected the result with the 2011 corn/soy binary layer. I repeated the process for new corn/soy lands as of 2012. I then multiplied these binary layers (1 = newly planted to corn/soy, 0 = previously or never planted to corn/soy) by the previous year’s land cover to determine which land uses had been displaced by new corn/soy. This analysis would have been sensitive to cumulative classification accuracies in the multiple time periods used, but was conservative in that it only identified locations that over the study period had never been classified as corn/soy prior to 2011 or 2012.


Long-term trends from statewide NASS statistics, 1980–2012

Statewide NASS crop statistics showed a long-term trend of increasing corn area planted in the Dakotas from 1980 through 2012. Regime shift detection using data from both states (USDA NASS 2013) showed that mean corn area planted averaged 17,313 km2 from 1980 until 1999, then experienced successive upward shifts in 2000, 2007, and 2012, reaching 39,457 km2 (Fig. 2a). The area planted to corn in the Dakotas increased from 5 to 10 % of all U.S. Cornland between 1980 and 2012 (USDA ERS 2013).
Fig. 2

Regime shift analysis of area planted to corn, soybeans, or wheat within North and South Dakota, 1980–2012, computed from statewide USDA statistics (USDA NASS 2013). Dashed line is the weighted regime mean (sensu Rodionov 2006). a Corn. b Soybeans. c Wheat

The two-state soybean area also increased substantially, with an abrupt upward regime shift in 1997, thought to be due to the release of cold- and drought-adapted high-yielding varieties, increased demand of international markets for soybean, easier management of glyphosate-resistant soybeans, and significant advantage to growers of output and net income compared with wheat (personal communication, Guo-Liang Jiang, Department of Plant Science, South Dakota State University, Brookings, SD). The area planted to soybeans was statistically constant for the next 15 years, but shifted upward again in 2012 (Fig. 2b). The ratio of soybean to corn area planted peaked at 1.4:1 in 2001, and in 2012 was ~1:1.

Wheat area declined as corn and soybean area increased. Wheat area planted was statistically stable at 57,753 km2 from 1980 until 2000 (with low and high outliers in 1983 and 1996, respectively), but shifted significantly downward in 2001 and 2011, culminating in a mean value of 40,373 km2 within the two states (Fig. 2c).

DPPR land use distribution, duration, and diversity from CDL data layers

Land cover in the DPPR is a mixture of crop and non-crop vegetation. A total of 67 CDL land-cover classes occurred within the DPPR during the period analyzed (Table 1). Non-crop cover types were grassland (including pasture and grass hayland), wetlands, water, and forest. Although widely distributed, non-crop lands were most abundant in ecoregions with topographic and/or soil limitations to cultivation (Fig. 1): the Missouri Coteau (subregions 42a, 42b, 42d), the Prairie Coteau Escarpment (46l), the Turtle Mountains (46b), and sandy glacial lake deltas (46d) and glacial outwash (46j).

Small grains and corn/soy tended to mirror each other in spatial distribution, with corn/soy concentrated in the southeast and small grains in the northwest portion of the region (Fig. 3). Although only 14 % of the DPPR was small grains as of 2012 (Table 1), 39 % of the DPPR had been planted to small grains at some point during the 7 years of CDL record, affecting a much larger geographic footprint due to crop rotation. Corn/soy was less mobile: 33 % of the DPPR was corn/soy as of 2012 (Table 1), and 46 % of the DPPR had been planted to corn/soy at some point during the 7 years.
Fig. 3

Distribution of major crops within the DPPR, computed from 2012 CDL data. a Percent of landscape planted to corn/soy, and the location of ethanol manufacturers b Percent of landscape planted to small grains

The variety of land uses mapped by CDL was least in the southeast portion of the DPPR, where landscape blocks (i.e., a 31.5 km2 aggregation of 187 × 187 pixels) were dominated by corn/soy, and greatest in the northwest because of the diverse crops grown there (Fig. 4). The least diverse landscape blocks were nearly continuous corn and soybeans, with only minor amounts of pasture/grassland, deciduous forest (i.e., farmstead shelterbelts), and rural development (i.e., roads and farm buildings). The most diverse landscape contained 35 different classes, including barley, buckwheat, sunflowers, canola, flaxseed, several types of pulses and small grains, alfalfa and other hays, as well as a variety of natural land types. It should be noted that this measure of landscape diversity is not equivalent to biodiversity, though, because the CDL classification has few categories of non-cropland.
Fig. 4

Diversity of CDL-mapped land uses within the DPPR

Nearly all of the region’s ethanol refineries were located in areas where corn/soy density was 50 % or more of the landscape (Fig. 3a). The only exception was the Underwood refinery, which is co-located with a power plant that provides steam for ethanol processing and drying distiller’s grains. The proportion of corn in the landscape was 34 % within 30 km of an ethanol plant (excluding Underwood), but only 13 % in areas beyond 30 km of an ethanol plant (Table 2).
Table 2

Percentage of land in major crops relative to proximity to an ethanol plant, based on 2012 CDL data


Land within 30 km of an ethanol plant (%)a

Land beyond 30 km of an ethanol plant (%)













aExcludes plant in Underwood, ND located on western border of study area

Short-term trends within the DPPR, 2006–2012

Analysis of the CDL-derived land use within the DPPR showed that crop trends over this shorter time period (2006–2012) and smaller land area were similar to those observed using the longer-term statewide statistics. There was a significant increase in corn/soy over seven years of data (Kendall’s tau = 0.905, p = 0.007), but the average rate of increase was nine times greater for 2010–2012 (+7,425 km2 year−1) than for 2006–2010 (+785 km2 year−1). The total area of corn/soy increased from 26.5 % of the region in 2006 to 35.5 % of the region by 2012.

The area of small grains decreased significantly over time (Kendall’s tau = −0.810, p = 0.016), an average loss of 1,926 km2 year−1 (Fig. 5). The area of “other crops” constituted about one-fourth the area of corn/soy, and was relatively constant throughout the period.
Fig. 5

Trends in major land use categories within the DPPR, 2006–2012, summarized from CDL data layers. Developed lands were masked out

Contrary to expectations, grassland area increased between 2006 and 2010, decreased slightly between 2010 and 2011, and then decreased more profoundly between 2011 and 2012 (Fig. 5). By 2012, grassland area was 9 % lower than it had been in 2006. There was no significant trend when the entire period was analyzed, but analysis of the 2006–2011 data showed a significant increase in grassland area (Kendall’s tau = 0.867, p = 0.024). Wetland area fluctuated, but exhibited no significant linear trend across the seven year period. Forest area was small and nearly constant (Fig. 5, bottom line).

Crop and non-crop areas approximately equaled each other, with no significant statistical trend in either value across the 7 year period. However, there was a large divergence that occurred between 2011 and 2012 as crop area increased by 16 % and non-crop area decreased by a comparable amount (Fig. 5, top two lines).

DPPR inter-annual land-use transitions from CDL data layers

Inter-annual transitions, 2006–2010

Inter-annual land-use transitions were fairly stable from 2006 through 2010, as described in this section, but inter-annual transitions after 2010 differed due to climate and accelerated agricultural expansion, described in subsequent sections. Inter-annual land-use transitions computed from the 2006–2010 CDL data layers showed that grassland, forest, and corn/soy had high self-replacement values, meaning that a pixel of that land-use type was highly likely to remain as that land-use type in the following year (Fig. 6). In the case of grassland and forest, self-replacement may indicate remnant natural vegetation. In the case of corn/soy, high self-replacement means that, once cropland becomes corn/soy, it is rarely rotated with other land uses.
Fig. 6

Box and arrow diagrams showing inter-annual land cover transition rates expressed as a fraction of the total area of the source pool, based on 2006 through 2012 CDL data layers. Values within boxes are self-replacement rates, whereas values positioned on arrows are land use changes. Dashed arrows = 0.02–0.15, thin solid arrows = 0.16–0.29, thick solid arrows ≥0.30. Land use change rates shown only for transitions exceeding 1750 km2

The inter-annual self-replacement rates for corn/soy were about 69 % between 2006 and 2010. The main transition out of corn was to small grains. A smaller fraction of corn/soy transitioned to grassland, presumably as fallow in a crop rotation. Transfers of corn/soy to grassland equaled or exceeded transfers of grassland to corn/soy during the first four transition periods.

Small grains were a more dynamic land-use type, with annual self-replacement rates of about 33 % during the first four time periods (Fig. 6). The main land uses that transitioned into small grains were corn/soy and other crops. Exchanges between small grains and grassland were bi-directional. Small grains had an equal probability (~27 %) of transitioning to corn/soy or other crops during the first four time periods.

Other crops were the most dynamic category, with annual self-replacement rates of only about 13 % during the first four time periods (Fig. 6). The major transition into and out of other crops was with small grains. Frequent rotations of crops such as sugar beets and sunflowers are implemented to reduce pests and disease and/or replenish the soil.

Grassland was the largest and most persistent land-use category between 2006 and 2010. Exchanges out of grassland into the other four categories were relatively constant at about 13,000 km2 year−1 during the first four transition periods, but exchanges into grassland were about 15,000 km2 year−1, which accounts for the gradual increase in area from 2006 to 2010 (Fig. 5).

Forest was the smallest land-use category (Fig. 5), with high inter-annual self-replacement rates. Exchanges into and out of forest were primarily with grassland and wetland/water (areas too small to be shown in Fig. 6), and were probably classification errors (e.g., confusion between #141 Deciduous Forest and #190 Woody Wetlands) or edge effects rather than true transitions.

Inter-annual wetland/water self-replacement rates during 2006–2010 were the most variable of any major land-use category, ranging from 54 to 72 % (Fig. 6). Grassland was the main exchange partner with wetland/water. The period from 2007 to 2008 exhibited the lowest wetland/water replacement value, the highest wetland-to-grassland transition rate, and a decrease in total wetland area (Figs. 5, 6), which was consistent with the drought conditions in North Dakota during that time (Appendix 1 in Supplementary material).

Effect of climate on 2010–2011 inter-annual transitions

As of 2011, the Northern Plains had experienced three successive springs of abnormally wet weather (Appendix 1 in Supplementary material). Additionally, late snowfalls during the spring of 2011 left a 10–25 cm snowpack over much of North Dakota as late as April 20 (Akyüz and Mullen 2011). The cold, wet weather prevented planting throughout much of northwestern North Dakota that year, and lands that had normally been small grains or other crops reverted to grassland and wetlands, resulting in gains in these covertypes (gray pixels, Fig. 7). These changes are evident in the unusually high transition rates to grassland from other crops (22 %) and small grains (24 %) in 2011. The wet conditions also explain the peak in non-crop area that occurred in 2011 (Fig. 5). There were losses of grassland to corn/soy in the central DPPR in 2010–2011 (black pixels, Fig. 7), but most of that grassland had been cropped in 2008, before the wet period began.
Fig. 7

Gains and losses of non-crop vegetation (grassland, forest, wetlands) and water in the DPPR, based on 2010–2011 CDL data

Effect of accelerated agricultural expansion on 2010–2012 inter-annual transitions

Agricultural expansion was greatest during 2010–2012. Corn/soy area increased by 27 %, consistent with the sharp increase reported by the USDA statewide statistics (Fig. 2), as corn price per bushel approximately doubled (USDA ERS 2013). Corn/soy self-replacement increased to 74 and 80 % in the last two transition periods, while small grain self-replacement decreased to 23 %. The proportion of corn/soy transitioning to small grains was 13 % in 2011–2012, as opposed to 19 % in 2006–2007. Small grains became much more likely to change to corn/soy in 2010–2011 (33 %) or 2011–2012 (45 %) than in the previous four transition periods.

The increase in corn/soy area also affected grasslands, both directly and indirectly via cascading effects on small grain/grassland exchanges. Transfers of grassland to corn/soy exceeded reciprocal transfers of corn/soy to grassland in 2010–2011 and 2011–2012, causing net annual gains of corn/soy (~4,000 km2 year−1). Exchanges between small grains and grassland became more unidirectional (grassland to small grains) in 2011–2012. The self-replacement fraction for grassland decreased from ~83 % during the first four transition periods to 75 % in 2011–2012, and in 2012 the area of corn/soy first exceeded the area of grassland (Fig. 5). The 2011–2012 transition was also the first of those analyzed in which there were substantial losses of wetlands to corn/soy, at 11 % of the 2011 wetland/water area.

Gains of corn/soy land between 2010 and 2011 were greatest in the center of the region, in the area that had experienced the largest loss of non-crop vegetation (black pixels, Fig. 7). The 2011–2012 gains in corn/soy land were more widespread, with most of the region exhibiting a 1–10 % gain (Appendix 2 in Supplementary material). Fifty-six percent of the 2012 corn/soy gain was from lands that had never been planted to corn/soy during the previous 6 years. Of these lands, 44 % had been wheat lands and 34 % had been grasslands in 2011 (Fig. 8). An equivalent area of never-before corn/soy land was converted to corn/soy between 2010 and 2011, with wheat and grasslands each contributing about 40 % to that new corn/soy land. Conversion of wetlands to new corn/soy land was relatively minor in 2011, but increased to 8 % of the new corn/soy land in 2012 because the drought that year (Appendix 1 in Supplementary material) facilitated wetland cultivation. A total of 10,215 km2 of new corn/soy land in 2011 and 2012 had never previously been planted to corn/soy within the study period.
Fig. 8

Land uses converted to new corn/soy land within the DPPR based on CDL data, 2010–2011 (total area = 5,057 km2) and 2011–2012 (total area = 5,158 km2)

Three-year land use rotations, 2006–2012

Twenty-four land use sequences accounted for 75–82 % of all three-year rotations (Table 3). Continuous forest (FFF) was the only major land use sequence in which forest occurred. Exchanges between grassland and wetlands/water (rows 13, 14) were due primarily to inter-annual variation in soil wetness, and are generally reversible. Grassland occurring only at the beginning of corn/soy sequence (GCC or GGC, row 11) may indicate the conversion of grasslands to corn/soy, and increased significantly over time. Grassland occurring in the middle year between two crops (CGC, SGS, OGS, SGO) was presumably hay or herbaceous fallow, but did not change significantly over time. Among the 24 top land use sequences, the only two combinations ending in grassland were continuous grassland or grassland alternating with wetland/water.
Table 3

Percent of all land area and trend (Kendall’s tau) for the 24 most common three-year land use sequences, based on CDL data


Land Use Sequence

2006–07–08 (%)

2007–08–09 (%)

2008–09–10 (%)

2009–10–11 (%)

2010–11–12 (%)

Kendall’s tau









































































































































C corn/soy, G grassland, O other crops, S small grains, W wetlands/water, F forest

*p < 0.05

Among crop rotations (rows 1–11, Table 3), continuous corn/soy (CCC, row 1) was the most common, accounting for 18.8 % of the region during 2010–11–12. The next most abundant rotations involved 2 years of small grains rotated with other crops (SSO, SOS, or OSS) or 2 years of corn/soy rotated with small grains (CCS, CSC, or SCC). The “other crops” were always in some combination with small grains. Rotations that decreased on the landscape all included 2–3 years of small grains (rows 5–6, Table 3).

Conversion of long-duration natural lands

Although the CDL does not distinguish native sod, areas that were continuously non-crop vegetation or water during the first five or 6 years of analysis, called long-duration natural lands, may serve as a proxy (Table 4). From 2010 to 2011, 4 % of the long-duration natural lands were converted to other land uses (about half to corn/soy). The loss rate was smaller from 2011 to 2012, both as a percent of initial (2.7 %) and a percent of the region (0.8 %), but a larger proportion was attributable to corn/soy expansion.
Table 4

Fate of long-duration natural land (water, wetland, grassland, forest) during 2010–2011 and 2011–2012, based on CDL data

Transition period

Initial area of long-duration natural land

Area lost to corn/soy

Area lost to all land uses

Area lost (% of initial)

Area lost (% of region)














Land-use changes due to agricultural expansion

In a shell game, a pea is placed under one of three walnut shells that are quickly shuffled, whereupon the player must guess its location. Land-use changes in the Northern Plains are like a shell game, because they are not unidirectional: multiple land uses are interchanged with each other, often in ways that are difficult to track. Unlike urban development or deforestation, which cause persistent landscape alteration in predictable locations, changes in agricultural practices can be much more ephemeral and reversible, rapidly responding to climate, markets, or policies.

In the 2 years between 2010 and 2012, corn/soy land in the DPPR expanded by 27 %, adding an area (+15,400 km2) larger than the U.S. state of Connecticut and half the size of the country of Belgium. Two-thirds of this land had never before been planted to corn/soy within the 2006–2012 study period, and one-sixth of the land previously had more natural cover. Thus, concerns about grassland loss are justified. However, the majority of the grassland impacted by agricultural expansion was not native prairie, but herbaceous vegetation incorporated into an agricultural land-use sequence either intentionally (as part of a planned rotation) or unintentionally due to unexpected climate extremes leading to crop failure.

Agricultural expansion affects multiple land uses, not just grasslands. Wheat and other small grains have been particularly impacted by corn/soy expansion, a fact overlooked by previous studies (Faber et al. 2012; Wright and Wimberly 2013). Small grain and corn/soy lands were historically separated by their different climate tolerances, but that separation is blurring as drought- and cold-resistant corn/soy varieties are developed, as evidenced by the region-wide corn/soy gain between 2011 and 2012 (Appendix 2 in Supplementary material).

The displacement of small grains by corn/soy has implications for food security. Wheat is directly consumed by humans, but field corn and soybeans generally are not. In 2012, 87 % of U.S. corn domestic corn use was for fuel production or livestock feed, the remainder going to “other food, seed, and industrial uses” (USDA ERS 2013). The displacement of small grains also has international implications, because about half of the U.S. wheat crop is exported, whereas only ~20 % of the U.S. corn crop is exported to other countries (USDA ERS 2013).

In addition to causing land-use replacements, agronomic changes in the DPPR are simplifying crop rotations. Crop rotation has long been recognized as a standard component of integrated pest management, but improvements in agricultural technology have enabled many producers to reduce crop rotation and rely solely on improved genetics and pesticides to control pests. However, simplifying crop rotations can and is providing an opportunity for weed, disease and insect organisms to build to harmful levels (Landis et al. 2008; Beck 2013).

Land-use policies and decision-making

Landscape-scale changes in agroecosystems result from the cumulative decisions of multiple land owners responding to available agricultural technologies, crop pricing, and policy. Understanding the interplay between ecological and social factors across multiple scales is integral to landscape change initiatives in productive agricultural regions (Atwell et al. 2009). The federal Conservation Reserve Program (CRP) encouraged farmers to convert environmentally sensitive land to grassland over 10- to 15-year contracts, and was quite popular in the DPPR. CRP had demonstrated benefits to soil carbon sequestration (McLauchlan et al. 2006) and wildlife (Feather et al. 1999; Reynolds et al. 2001). However, CRP payments are currently much less than cash rents or crop profits, and it is anticipated that many farmers have or will put CRP lands back into crop production as existing contracts expire.

The Renewable Fuel Standard in the U.S. Energy Independence and Security Act of 2007 (EISA) has spurred agricultural expansion in the DPPR by promoting production of biofuels. Recognizing the conflict between use of crops for food versus fuel, EISA specifies that 58 % of the mandated 2022 ethanol total be must be derived from non-cornstarch products, such as cellulose. However, commercial production of cellulosic ethanol is still elusive. Switchgrass (CDL class 60), heralded as a future cellulosic ethanol crop, was planted on only 2.4 km2 in the DPPR in 2012.

Adoption of more sustainable agricultural practices entails moving toward integrated agricultural systems and offering incentives or imposing regulations to affect farmer behavior (Dale et al. 2013). Valuation of the social benefits provided by sustainable agriculture may be one way to recognize and potentially compensate farmers for their best management practices (Polasky et al. 2011).

Future land-use change

The work presented here constitutes the first two steps in the framework proposed by Houet et al. (2010) to explore subtle land-use and land-cover changes: (1) study site definition and (2) system analysis (trends analysis, landscape dynamics, seeds of future change). I did not elaborate nor assess future land-use scenarios, but others have done so. A simulation of converting row crops to biofuel switchgrass in the James River valley of the central DPPR demonstrated water quality improvement (Wu et al. 2012), and a simulation of land use changes to 2020 driven by biofuel mandates was consistent with my findings on recent trends: gain of corn/soy and loss of wheat and CRP land in the Northern Plains (Mehaffey et al. 2011).

Climate change is a wild card in forecasting agricultural land uses for the DPPR. This analysis demonstrated extensive temporary abandonment of cropland in 2011 due to a climate extreme of excessive wetness, but the region is equally susceptible to drought. The 2012 U.S. Midwest drought decreased average per-area corn yields by 29 % in South Dakota (USDA NASS 2012). The 2012 drought also demonstrated that diverse rotations reduce risk associated with adverse weather, because the critical water uptake period for wheat occurs earlier in the growing season than for corn or soybeans, and wheat producers harvested an above-average crop in 2012 (Beck 2013). The extreme drought of the 1930s caused the greatest climatic impact to date: the land planted to corn in South Dakota dropped from 22,258 km2 in 1931 to only 12,214 km2 a decade later (USDA NASS 2013), illustrating how quickly land-use reversals could occur under prolonged severe drought. The 2012 area planted to corn in South Dakota was 24,888 km2 (USDA NASS 2013).

Crop pests may also be a wild card in the future sustainability of corn/soy. The western corn rootworm (Diabrotica virgifera) has been kept at bay by corn/soy rotations and transgenic Bt corn hybrids, but is quickly adapting and developing resistance to these control methods (Gray et al. 2009; Beck 2013). Corn rootworm could be a much greater threat to the sustainability of the crop in the future (personal communication, Jonathan Lundgren, USDA Agricultural Research Service, Brookings, SD). Weed resistance to the widely used herbicide glyphosate is also increasing in areas planted to transgenic glyphosate-resistant corn and soybeans (Powles 2008).

The use of the term “shell game” typically implies that someone is being conned by the person shuffling the shells. No such subterfuge is implied by this paper, and there is no “con man” in this game. However, the use of terms such as “marginal lands” to describe potential new sites for energy crops is misleading, because all land uses have value. When one land use expands, it is always at the expense of another. Sustainable landscapes in the DPPR require an understanding of the land-use costs of agricultural expansion, as well as their social and environmental consequences.



Lisa Schulte and two anonymous reviewers are thanked for comments that improved the quality of this manuscript.

Supplementary material

10980_2013_9947_MOESM1_ESM.pdf (10 kb)
Supplementary material 1 (PDF 11 kb) Appendix 1. Palmer Hydrological Drought Index for North and South Dakota, March single month values for 2004-2013 (NCDC 2013)
10980_2013_9947_MOESM2_ESM.tif (3.4 mb)
Supplementary material 2 (TIFF 3483 kb) Appendix 2. Location of gains in corn/soy land within the DPPR, based on 2011-2012 CDL data. Values are the proportion of land area that became new corn/soy. Values ≤ 0 lost corn/soy land or were unchanged; positive values gained corn/soy land


  1. Akyüz A, Mullen BA (2011) 2011 growing season weather summary for North Dakota. North Dakota State Climate Office, FargoGoogle Scholar
  2. Atwell RC, Schulte LA, Westphal LM (2009) Landscape, community, countryside: linking biophysical and social scales in US Corn Belt agricultural landscapes. Landscape Ecol 24(6):791–806CrossRefGoogle Scholar
  3. Beck R (2013) Benefits of wheat in a corn soybean rotation. South Dakota State University, Brookings, SD. Accessed Aug 2013
  4. Boryan C, Yang Z, Mueller R, Craig M (2011) Monitoring US agriculture: the US Department of Agriculture, National Agricultural Statistics Service, Cropland Data Layer Program. Geocarto International 26(5):341–358Google Scholar
  5. U.S. Census Bureau (2010) State and County QuickFacts. U.S. Department of Commerce, Washington DC. Accessed Aug 2013
  6. Culman SW, Young-Mathews A, Hollander AD et al (2010) Biodiversity is associated with indicators of soil ecosystem functions over a landscape gradient of agricultural intensification. Landscape Ecol 25(9):1333–1348CrossRefGoogle Scholar
  7. Cunningham MA (2006) Accuracy assessment of digitized and classified land cover data for wildlife habitat. Landscape Urban Plan 78(3):217–228CrossRefGoogle Scholar
  8. Cunningham MA, Johnson DH (2006) Proximate and landscape factors influence grassland bird distributions. Ecol Appl 16(3):1062–1075PubMedCrossRefGoogle Scholar
  9. Dale VH, Kline KL, Kaffka SR, Langeveld JWA (2013) A landscape perspective on sustainability of agricultural systems. Landscape Ecol 28(6):1111–1123CrossRefGoogle Scholar
  10. Faber S, Rundquist S, Male T (2012) Plowed under: how crop subsidies contribute to massive habitat losses. Environmental Working Group, Washington DCGoogle Scholar
  11. Fang S, Gertner G, Wang G, Anderson A (2006) The impact of misclassification in land use maps in the prediction of landscape dynamics. Landscape Ecol 21(2):233–242CrossRefGoogle Scholar
  12. Feather P, Hellerstein D, Hansen L (1999) Economic valuation of environmental benefits and the targeting of conservation programs: the case of the CRP. Agricultural Economic Report No. AER-778, U.S. Department of Agriculture Economic Research Service, Agricultural Economic Report No. AER-778, Washington, DC, USAGoogle Scholar
  13. U.S. GAO (2007) Agricultural conservation: farm program payments are an important factor in landowners’ decisions to convert grassland to cropland. GAO-07-1054 Report to Congressional Requesters, United States Government Accountability Office, Washington DCGoogle Scholar
  14. Gray ME, Sappington TW, Miller NJ, Moeser J, Bohn MO (2009) Adaptation and invasiveness of western corn rootworm: intensifying research on a worsening pest. Annu Rev Entomol 54(1):303–321PubMedCrossRefGoogle Scholar
  15. Houet T, Loveland TR, Hubert-Moy L et al (2010) Exploring subtle land use and land cover changes: a framework for future landscape studies. Landscape Ecol 25(2):249–266CrossRefGoogle Scholar
  16. Johnston CA (2013) Wetland losses due to row crop expansion in the Dakota Prairie Pothole Region. Wetlands 33(1):175–182CrossRefGoogle Scholar
  17. Landis DA, Gardiner MM, van der Werf W, Swinton SM (2008) Increasing corn for biofuel production reduces biocontrol services in agricultural landscapes. Proc Natl Acad Sci 105(51):20552–20557PubMedCrossRefGoogle Scholar
  18. Langpap C, Wu J (2011) Potential environmental impacts of increased reliance on corn-based bioenergy. Environ Resour Econ 49(2):147–171CrossRefGoogle Scholar
  19. McLauchlan KK, Hobbie SE, Post WM (2006) Conversion from agriculture to grassland builds soil organic matter on decadal timescales. Ecol Appl 16(1):143–153PubMedCrossRefGoogle Scholar
  20. Mehaffey M, Smith E, Van Remortel R (2011) Midwest U.S. landscape change to 2020 driven by biofuel mandates. Ecol Appl 22(1):8–19CrossRefGoogle Scholar
  21. National Research Council (2003) Frontiers in agricultural research: food, health, environment, and communities. National Academy Press, Washington DCGoogle Scholar
  22. NCDC (2013) Plot time series. National Climatic Data Center, NOAA, Asheville, NC. Accessed Aug 2013
  23. NDSU (1997) Corn production guide. NDSU Extension Service, North Dakota State University of Agriculture and Applied Science, FargoGoogle Scholar
  24. Polasky S, Nelson E, Pennington D, Johnson K (2011) The impact of land-use change on ecosystem services, biodiversity and returns to landowners: a case study in the State of Minnesota. Environ Resour Econ 48(2):219–242CrossRefGoogle Scholar
  25. Powles SB (2008) Evolved glyphosate-resistant weeds around the world: lessons to be learnt. Pest Manag Sci 64(4):360–365PubMedCrossRefGoogle Scholar
  26. R Core Team (2013) R: a language and environment for statistical computing. R Foundation for Statistical Computing, ViennaGoogle Scholar
  27. Ransom J (2004) Basics of corn production in North Dakota. A-834, NDSU Extension Service, North Dakota State University of Agriculture and Applied Science, FargoGoogle Scholar
  28. Reynolds RE, Shaffer TL, Renner RW, Newton WE, Batt BDJ (2001) Impact of the Conservation Reserve Program on duck recruitment in the U.S. Prairie Pothole Region. J Wildl Manag 65(4):765–780CrossRefGoogle Scholar
  29. Riseng CM, Wiley MJ, Black RW, Munn MD (2011) Impacts of agricultural land use on biological integrity: a causal analysis. Ecol Appl 21(8):3128–3146CrossRefGoogle Scholar
  30. Rodionov SN (2006) Sequential regime shift detection, version 3.2 (software add-into Excel 2002). NOAA Pacific Marine Environmental Laboratory, Seattle, WA, USA. Accessed Aug 2013
  31. Samson FB, Knopf FL, Ostlie WR (2004) Great Plains ecosystems: past, present, and future. Wildl Soc Bull 32(1):6–15CrossRefGoogle Scholar
  32. Searchinger T, Heimlich R, Houghton RA et al (2008) Use of U.S. croplands for biofuels increases greenhouse gases through emissions from land-use change. Science 319(5867):1238–1240PubMedCrossRefGoogle Scholar
  33. Secchi S, Gassman PW, Jha M, Kurkalova L, Kling CL (2011) Potential water quality changes due to corn expansion in the Upper Mississippi River Basin. Ecol Appl 21(4):1068–1084PubMedCrossRefGoogle Scholar
  34. Shao G, Wu J (2008) On the accuracy of landscape pattern analysis using remote sensing data. Landscape Ecol 23(5):505–511CrossRefGoogle Scholar
  35. Stephens SE, Rotella JJ, Lindberg MS, Taper ML, Ringelman JK (2005) Duck nest survival in the Missouri Coteau of North Dakota: landscape effects at multiple spatial scales. Ecol Appl 15(6):2137–2149CrossRefGoogle Scholar
  36. Stephens SE, Walker JA, Blunck DR et al (2008) Predicting risk of habitat conversion in native temperate grasslands. Conserv Biol 22(5):1320–1330PubMedCrossRefGoogle Scholar
  37. Tan ZX, Liu SG, Johnston CA et al (2005) Soil organic carbon dynamics as related to land use history in the northwestern Great Plains. Glob Biogeochem Cycles 19(3):GB3011CrossRefGoogle Scholar
  38. Tilman D, Balzer C, Hill J, Befort BL (2011) Global food demand and the sustainable intensification of agriculture. Proc Natl Acad Sci 108(50):20260–20264PubMedCrossRefGoogle Scholar
  39. U.S. Energy Information Administration (2013) South Dakota state energy profile. Accessed Aug 2013
  40. USDA NASS (2012) Crop Production (November 2012). USDA, National Agricultural Statistics Service, Washington DCGoogle Scholar
  41. USDA ERS (2013) Crops. U.S. Department of Agriculture, Economic Research Service, Washington DC, USA. Accessed Aug 2013
  42. USDA NASS (2013) Statistics by state. U.S. Department of Agriculture, National Agricultural Statistics Service, Washington DC. Accessed Aug 2013
  43. Vaché KB, Eilers JM, Santelmann MV (2002) Water quality modeling of alternative agricultural scenarios in the U.S. corn belt. JAWRA. J Am Water Resour Assoc 38(3):773–787CrossRefGoogle Scholar
  44. Wallander S, Claassen R, Nickerson C (2011) The ethanol decade: an expansion of U.S. corn production, 2000–2009. EIB-79, USDA Economic Research Service, Washington DCGoogle Scholar
  45. West TO, Post WM (2002) Soil organic carbon sequestration rates by tillage and crop rotation. Soil Sci Soc Am J 66(6):1930–1946CrossRefGoogle Scholar
  46. Wright CK, Wimberly MC (2013) Recent land use change in the Western Corn Belt threatens grasslands and wetlands. Proc Natl Acad Sci 110(10):4134–4139PubMedCrossRefGoogle Scholar
  47. Wu Y, Liu S, Li Z (2012) Identifying potential areas for biofuel production and evaluating the environmental effects: a case study of the James River Basin in the Midwestern United States. GCB Bioenergy 4(6):875–888CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2013

Authors and Affiliations

  1. 1.Department of Natural Resource ManagementSouth Dakota State UniversityBrookingsUSA

Personalised recommendations