Land Use Change and Policy in Iowa ’ s Loess Hills

Land use changes have important implications on ecosystems and society. Detailed identification of the nature of land use changes in any local region is critical for policy design. In this paper, we quantify land use change in Iowa’s Loess Hills ecoregion, which contains much of the state’s remaining prairie grasslands. We employ two distinct panel datasets, the National Resource Inventory data and multi-year Cropland Data Layers, that allow us to characterize spatially-explicit land use change in the region over the period 1982-2010. We analyze land use trends, land use transitions and crop rotations within the ecoregion, and contrast these with county and state-level changes. To better comprehend the underlying land use changes, we evaluate our land use characterizing metrics conditional on soil quality variables such as slope and erodibility. We also consider the role of contemporary agricultural policy and commodity markets to seek explanations for land use changes during the period of our study. Although crop production has expanded on the Loess Hills landform since 2005, much of the expansion in corn acres has been from reduced soybean acreage. We find that out of the total 258 km2 increase in corn acreage during 2005-’10, about 100 km2 transitioned from soybeans. Data also indicate intensifying monoculture with higher percentage of corn plantings for two to four consecutive years during 2000-’10. In addition, crop production is found to have moved away from more heavily sloped land. Cropping does not appear to have increased on lands with higher crop productivity.

highest benefit per unit cost. Secchi et al. (2010) used the Cropland Data Layer (CDL) to simulate the extent to which biofuels-related expansion may tilt Iowa crop rotations toward more corn intensive rotations. Brown and Schulte (2011) studied aerial photographs to document the decline of small grains and grass agriculture in three Iowa townships between 1937 and 2002. Miller (2006) commented on the roles of urban pressure, topography, erodibility constraints and agro-economic incentives on assembling a remnant prairie, the Broken Kettle Grassland Preserve, at the ILHL's north end. Many technical contributions to our understanding of soil and water conservation on the landform have also been published (e.g., Tomer et al. 2007).
Most relevant to our study, Farnsworth et al. (2010) connected privately obtained, remotely sensed data on land cover with crop productivity information to develop a conservation priority index that also seeks to account for benefits from tract connectivity. Their inquiry was static with 2006 land uses, just before major changes in United States cropping activities. Arora et al. (2015) used CDL data to quantify land use transitions in the ILHL between 2001 and 2013. They found that grass acres had declined during this period in the ILHL, having moved into wooded categories, and also that corn acres had expanded largely at the expense of soybean acres. They expressed surprise, however, at the limited expansion of row-crop production in the region.
This study seeks to provide a more detailed scrutiny of recent land use change across the ILHL. We provide an overview of relevant policies and the evolving market environment, followed by an explanation of our materials and methods; primarily different land use data sources and data processing procedures. After analyzing results, we summarize land use conversion trends with a brief discussion.

Policy and Market Environment
The past thirty years has seen a shift in the emphasis of United States agricultural policy away from food and feed production toward energy outputs and also toward environmental outputs that are not generally supported through market incentives. The main agricultural policies of relevance have been those regarding conservation, biofuels and crop insurance.  The 1985 Farm Bill also saw the introduction of conservation compliance provisions whereby those who farm highly erodible lands may be ineligible for some forms of agricultural income support. Growers planting on highly erodible land commit to a conservation plan in order to become compliant. Between 1996 and 2014, eligibility for crop insurance premiums was not conditioned on conservation compliance, but linkage was re-established under the 2014 Farm Bill. Recent trends in cropping systems, to be discussed later, have made compliance easier than was the case before the mid-1990s.
For decades preceding the 1996 Farm Bill, the commodity-specific income support that growers received depended in large part on cropping choices (Novak et al. 2015). Some crops, collectively labeled 'program' crops, received support in proportion to acres and yields. Corn was a program crop but soybeans and grass/hay were not. The de-linking of cropping choices and subsidies in the 1996 Farm Bill was motivated by the costs of inflexibility in marketplace response (as growers would lose non-market support upon adapting to market prices) and by International Trade Agreement commitments.
Crop insurance has had at least some federal support since the 1930s, but was not seen as an integral component of income support until the 1990s (Glauber 2013). In an effort to promote program performance by expanding participation, commencing in 1994 a series of legislative ________________________________________________ enactments increased premium subsidies and expanded contract choices. Upon passage of the 2014 Farm Bill, crop insurance support had become a firmly established primary pillar of agricultural income support. Although pre-subsidy rates are required to be actuarially fair, as far as is practical, the U.S. Government Accountability Office (U.S. GAO 2015) has discerned underpricing in production-riskier counties, however, none of these are in Iowa. Nonetheless, the growth of crop insurance subsidies, unavailable or less generous for grass-based activities, is likely to promote crop production (Claassen et al. 2011;Feng et al. 2012;Miao et al. 2014).
Corn-based ethanol has been, indirectly or directly, promoted by the U.S. federal government since the 1970s. Direct support for ethanol production as a renewable fuel came through federal Other commodities also saw price increases so that they remain a ________________________________________________ competitive use of land resources. A secondary effect was through beef markets. Farm-level beef prices rose in part because feedlot owners needed to cover higher corn input costs or go out of business. The U.S. national beef herd has declined over the 1996-2014 period in the face of higher feed input prices and also adverse weather conditions, with a sharp decline after 2008. Even so, grassland rental prices also increased as they provide an alternative to corn-based cattle feed.
Marked technological change has occurred in crop production during recent decades. Perhaps most relevant to this study are expanded use of conservation tillage and the advent of genetically modified corn and soybean seeds. Conservation tillage can reduce production costs and can also preserve moisture as a risk management strategy against drought. The Loess Hills area, though far from arid, is among the driest in Iowa. Conservation tillage also protects against soil erosion and is viewed as an acceptable strategy for conservation compliance. However, tillage also provides weed control (Carpenter and Gianessi 1999), such that growers had been reluctant to adopt conservation tillage. The advent of glyphosate tolerant seeds allowed for cost-effective weed control by use of a single chemical after planting. There is substantial evidence that glyphosate tolerant seed complements less intensive tillage (Perry et al. 2015). In reducing the cost of conservation compliance, these seeds may facilitate corn and soybean production on erodible land. Growers in the SCA have been early and extensive adopters of both conservation tillage and glyphosate tolerant seed. 4 ________________________________________________ 4 Over the 1998-2001 period 47% of soybean growers in the SCA adopted both conservation tillage and glyphosate tolerant soybean seed, when compared with 37.4% nationwide. For the 2007-2011 period the comparable figures were 72.5% and 65.3%. See table 1 and supporting text in Perry et al. (2015) for explanations of data. Ed Perry kindly made SCA summary data available to us.

Materials and Methods
To address land use changes in Iowa's Loess Hills and related policy implications, we conducted a comprehensive data analysis at two levels of aggregation: SCA and ILHL. We used two data sources: National Resource Inventory (NRI) data and CDL data. NRI data allow for evaluation of historical land use changes at the county level of aggregation in the SCA , where CDL data are only available for recent years (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014). On the other hand, and in contrast with NRI data, CDL data are spatially-delineated and so allow us to evaluate changes specific to ILHL.

National Resource Inventory Data
We utilized National Resource Inventory (NRI 2013) data to evaluate land use/land cover (LULC) trends for the SCA encompassing the ILHL. Focusing on point-level data, the NRI is a survey-based longitudinal database that provides comprehensive information on land characteristics as well as historical uses. In order to conform with USDA confidentiality protocols, and unlike the CDL, NRI data suppress spatial geo-coordinates (NRI 2013) although county location is provided. NRI data collection is based on a robust survey methodology that assures reliable and temporally-consistent estimates of land use and land quality parameters.
Specifically, the included sample points are intended to represent the overall geographic spread and heterogeneity of natural resources at national and regional levels (Nusser et al. 1998). Each sample point is accompanied by its representative weight measured in 100 acre (0.405 km 2 ) units that differ across sample points. A total of 2,600 NRI points span the SCA's 12,985 km 2 area.
Data range from 1982 to 2010, where each NRI point was observed every five years from 1982 to 1997 but annually commencing 2000.
The NRI dataset provides land quality parameters such as land capability classification (LCC) and erodibility index (EI). LCC groups soils into eight classes with regards to limitations for cropping, with higher class codes referring to more severe limitations. LCC classes I and II are most suitable for cropping whereas classes III and IV are more limited in use. Higher LCC classes have severe limitations and are considered to be unsuitable for cultivation (Helms 1992, p.65). EI measures soil's erosion potential, whereby soils with higher index values are costlier for cropping, as they entail pertinent management costs to limit erosion and preserve crop productivity (NRI 2013). About 1% of NRI points had non-constant LCC and/or EI values during the 1982-2010 period. These points were excluded when assessing land use by LCC and/or EI status. Within this domain of constant LCC lands; 6.6% of points have missing LCC for all years, 40.2% were in classes I-II, 48.3% were in classes III-IV, and 4.9% were assigned higher classes.
Locations with EI values that were either missing or temporally varying accounted for 31.1% of NRI points, with a disproportionate fraction in the urban category. Among the non-missing, constant EI points, the index ranges from 0.8 to 160.2 where 32.2% had EI ≥ 8. The USDA's Natural Resource Conservation Service (NRCS) defines soils with EI ≥ 8 as highly erodible.
Since the 1996 Farm Bill, soils with EI ≥ 8 have been eligible for CRP regardless of other attributes.

Cropland Data Layers
Pre-processing: Spatially explicit raster CDL data were downloaded from the CropScape portal of the National Agricultural Statistics Service (USDA NASS 2012) and clipped by the SCA encompassing the ILHL (figure 1). These CDL data are produced annually for the 48 contiguous states (since 2000 for Iowa) using a combination of (1) multiple satellite imagery dates each year to capture crop phenology differences, (2) concurrent USDA FSA training/validation data, and Downloaded CDL data could not be analyzed directly due to inherent differences in (1)  Also, spatial inconsistencies within the 'Fallow/Idle Cropland' class through time have been identified in prior studies (Kline et al. 2013). Laingen (2015) has recommended that all data be treated with circumspection, placing emphasis on the data generation processes. Once CDL classes were equivalent among the {2001, 2005, 2010, 2013} time steps, we then used a matrix union function to combine the information into one physical image. This matrix union function, when applied to two consecutive CDL years (e.g., Y1-Y2), provides a "from-to" attribute in the resulting thematic output attribute table with which one may track land cover changes between those years. Here, we simply applied this function three consecutive times (Y1 -Y2, Y1-2 -Y3, and finally Y1-2-3 -Y4) to achieve the four-year (Y1-2-3-4) "from-to" vector of change for each pixel. A 3x3 pixel majority despeckling function was then applied to this resulting image to simultaneously weed out spurious and illogical single pixels of changes through time (e.g., water-corn-forest-soy). After the multi-temporal despeckling operation was complete, the four-year "from-to" change vector attribute was then used to guide recoding of the single, thematic image back into the four respective CDL years as separate, thematic, image layers.
To allow for further consideration on the sorts of land that have seen changing use, we have linked CDL data to land quality. Corn Suitability Ratings (CSR, Miller 2005)

Analysis of Land Use Change
We present land use trends and land use transition matrices to characterize conditional (on land quality) and unconditional land use changes for SCA and ILHL. We seek comparisons among the two independent datasets, as well as among the two aggregation levels using unconditional land use trends. Last, but not least, to facilitate further scrutiny we empirically specify the structure of rotations for corn over the 2000-2010 period to assess whether cropping in this region is moving toward monoculture. We present our findings below.

Unconditional Land Use Trends
To utilize the full extent of NRI data while emphasizing more recent available data , table 2 provides summary data on cultivated area under crops, hayland and pastureland for 1982, 1992, 2001, 2005  about 100 km 2 so that, perhaps surprisingly, there has been no net change in total acres devoted to row crops over the period. There has been expansion of the forest (+67 km 2 ) and fallow/idle cropland (+70 km 2 ) categories. The ILHL is topographically variable, especially in regard to slope and land quality. Within the region there have been some subtle changes. Crop production ________________________________________________ 7 High erodibility is the only eligibility criterion for the CRP general sign-up. The continuous sign-up, introduced in the 1996 Farm Bill, also targets land that adopt certain conservation practices, such as wetland restoration and conserving riparian buffers. Continuous sign-up contracts with high priority conservation practices can be enrolled any time during the year. The NRI records CRP lands under continuous sign-up in their respective categories like cropland, forest, grassland, etc. (NRI 2013, p.12). 8 Perceptions among concerned observers in the area are that developed areas are expanding more rapidly than these data would suggest (personal correspondence with Susan Hickey, The Nature Conservancy). has expanded toward the north, especially in Woodbury and Monona counties where cropping is extensive. Crop production has declined in the more southerly counties where the landform is thin and cropping is limited. Crop production has generally moved from soybeans and toward corn overall along the hills but corn has not expanded in Plymouth County to the north or Because CDL classification protocols for developed acres have evolved substantially through these years, CDL data do not directly allow for an assessment of change to this category. In a separate query that appropriately adjusted for the redefinitions, we found a 2.6% increase in ILHL development acres (from 103 km 2 acres to 106 km 2 ) over the 2001-3.13 period. We had expected a larger increase.

Conditional Land Use Trends
Seven-County-Area: Table 4 presents land use trends conditional on being in LCC classes I-II and on being in classes III-IV. Classes I-IV contain almost all corn and soybean acres in this region. Furthermore, 75% or more of hay and pasture acres also lie in LCC classes I-IV.
However, relatively higher hay, pasture, forests and CRP land use shares are found on LCC classes III-IV. Corn acres have seen a slight migration to better land, perhaps because of incentives that the CRP provides to more limited land, while hay and pasture acres have shifted away from LCC classes I-II. land with EI ≥ 8. Also, only 13%-23% of total hayland and pastureland were present over the years considered for analysis. Acres under hay, pasture and CRP categories generally had higher erodibility index values. It is noteworthy that corn acres on EI ≥ 8 land was lower in 2010 when compared with its 1982 counterpart. This outcome may be due in part to the advent of CRP, and in part to conservation compliance constraints. Table 6 shows that corn presence and recent corn expansion have been concentrated on less steeply sloped river valley tracts that cut through the landform, see figure 5. Total acres to corn and soybeans on slopes >10% has changed minimally over the period. The moderate decline in the hay/pasture/grass category over 2001-2013 occurred mainly on shallow slopes while expansion in forest occurred throughout. Expansion in the fallow/idle cropland category occurred mainly on lower (≤ 5%) and higher slopes (> 15%) where the highest proportional expansion of this category has been on higher slopes, a notable observation given that overall CRP acres on the landform likely declined over the 2001-2013 period. Table 7 reports land use change by four corn suitability rating (CSR) categories. These are ≤ 69, or least suitable, 70-79, 80-89 and ≥90, see Miller (2005). Corn expanded and soybean contracted in each category over the 2001-2013 period. Total land in either corn or soybeans increased slightly on lower quality land but decreased slightly on better quality land, a perplexing finding that also applies for the SCA (table not shown). Land area under forest has expanded or trended sideways for all land quality categories while the data indicate that area in the hay/pasture/grass category has declined slightly for better land categories. Fallow/idle cropland is found to have expanded on higher quality land and to have contracted on lower quality land, perhaps due to high error in the category (as previously mentioned) or forest encroachment.

Temporal Transition Matrices
We also present temporal transition matrices for the SCA using NRI data as well as CDL land cover data. Pivot tables such as table 8 provide a matrix of land use transitions over time. They allow us to identify interesting transitions among the land use categories under study, which may have important policy implications for this region. We analyze pivot tables using NRI data for -2005, 2005and using CDL data for 2001-2005, 2005 periods. Whereas CDL data allows us to capture the most recent transitions, NRI data helps in quantifying conversions into urbanization. consistent with the national movement toward more cropland to meet growing demand for commodities. The shift back to hay/pasture/grass and fallow/idle cropland over 2010-2013 as well as contraction in both corn and soybean acres is not consistent with national trends. In 2013, corn and soybean area planted in the United States were, respectively, 384,000 km 2 and 308,000 km 2 . In 2010 the corresponding numbers were 348,000 km 2 and 312,000 km 2 . Many of the additional corn acres have come from outside the traditional Cornbelt, especially from Great Plains states such as North Dakota and Kansas. Table 8 shows that much of the corn acreage that moved out of cropping between 2001 and 2005 in the ILHL likely went into grass and fallow cropland, a pattern that was reversed in the subsequent five years so that corn acres were 138 km 2 larger in 2010 than in 2001. High corn acreage was sustained in the 2010-2013 period through declining soybean acres. While shifts occurred into grass and fallow cropland categories, net grass acres have declined due to outward transitions into the forests category (not included here) where invasive eastern red cedar is a problem in the area. Table 9

Structure of Rotations
As all indicators hold that corn acreage has expanded in the region dominated by corn-soybean rotations while cropped acreage has seen very limited change, it is certain that more corn intensive rotations are being used. Table 10 provides confirmation using CDL data for the ILHL.
The table shows that the percent of all corn land in a given year that returns to corn the next year has trended upward over the years. The pattern is more obvious when three and four year corn sequences are viewed. Figure 6 depicts CDL data that provide evidence of more intensive corn rotations toward the landforms thick north end. Our finding corroborates Plourde et al. (2013), who used CDL data across much of the Greater Mississippi watershed to discern an intensification of corn in rotations during 2003-2010.

Discussion
Several factors have led to growth in tilled acres across the United States since 2006. In the Loess Hills we conclude that corn production has increased where much of the expansion has been through displacement of soybeans. The evidence does point to grassland loss in the area where we remind the reader that pixel-level misclassification rates is very high for CDL data on grass and fallow/idle categories. Tables 2 and 3 show that the hay/pasture/grass category declined by about 12% and 17%, respectively, over 2005-2013 while the respective figures for the fallow/idle category are declines of about 15% and 21%. Even if evidence on grassland loss is discarded, there are adverse implications for environmental services as corn production in rotation is believed to improve soil quality (Karlen et al. 2006) as well as reduce demand for chemicals that improve fertility (Stanger and Lauer 2008) and manage pests (Gassmann et al. 2014).
Some parts of the ecoregion have seen cropland expansion, most notably southeast of Sioux City, while crop production has declined toward the less heavily cropped south. Both corn and forest acres have expanded everywhere in the hills over 2005-2013 but it should be noted that separating forested land and grass cover is problematic, especially in the presence of invasive shrubs. There is little evidence that cropping has moved to better quality land although there is some evidence that it has moved away from steeper slopes. The limited evidence available does not point to urban development as a major factor in the area but our view is that the matter warrants further inquiry. Land identified as fallow/idle has declined, likely due to a net decline in CRP acres.
Given the various forces that have aligned in recent years to incentivize row crop production and given national trends, why row cropping has not expanded by more along the hills is unclear.
Perhaps, for some reason, trends toward mechanization have favored less hilly land. Perhaps too, notwithstanding the growth in reduced tillage methods throughout the United States, conservation compliance regulations and targeted CRP sign-ups have proven to be more effective in protecting grass and wooded land in the area than elsewhere? Whether the Loess Hills are distinctive or our finding reflects a more general pattern of comparative constraint in row crop activity on hill terrain in recent years is a matter that warrants further inquiry.    Source: National Resource Inventory data.