Multi-Scalar Drivers and Patterns of Abandoned Cropland in the United States: Implications for Renewable Energy
Open Access
- Author:
- Baxter, Ryan
- Graduate Program:
- Geography
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- December 08, 2023
- Committee Members:
- Brian King, Program Head/Chair
Tom Richard, Outside Unit & Field Member
Douglas Miller, Major Field Member
Erica Smithwick, Chair & Dissertation Advisor
Helen Greatrex, Major Field Member - Keywords:
- Energy
Landuse
Landscape
GIS
Cropland
Landscape Ecology
Agriculture
Geography
Spatial Modeling - Abstract:
- Scaling up renewable energy production, solar and biomass, will require a substantial area of land. Such land should ideally not be in competition with other uses such as agriculture, pasture, or urban development. Furthermore, converting this land to renewable energy should neither contribute to carbon emissions nor negatively impact habitat or ecosystem services. Recently abandoned cropland that has not converted to another use like urban development or pasture is presented here as an appropriate land opportunity. This dissertation investigates abandoned cropland in three ways. First, it estimates how much usable abandoned cropland (UACL) there is in the United States. Usable abandoned cropland is an area recently in agricultural production but is not anymore and has not converted to another land use like urban development or pasture. Data from the Census of Agriculture, supplemented by information derived from the National Land Cover Dataset (NLCD), are used to calculate the amount of UACL that has emerged in the counties in the conterminous United States. Results show that between 5 and 15Mha of UACL exists in the United States, considering historical limits for when cropland was abandoned of 2007 and 1978 respectively. The relative contributions of these estimates to national energy consumption amount to only 3 to 8 percent of total light duty gasoline consumption, and make more significant contributions to other end uses, such as aviation. Second, this dissertation explores the potential drivers of cropland abandonment. Random forest analysis is performed to see which of a suite of 22 variables of underlying characteristics is important to variation in levels of UACL in each county in the eastern United States. The western United States is excluded due to the prevalence of irrigation, which is a stark difference from the east and would complicate analysis. Since many counties do not have a net increase in UACL, first, a classification analysis using Random Forest is performed to identify variables important to whether a county has UACL or not. Second, a regression analysis using Random Forest is performed to identify which variables are most important to the amount of net UACL that has emerged in each county that experiences an increase. Results show that the rent farmers pay to operate cropland and the value of farmland and buildings are among the most important variables influencing UACL. Interestingly, when viewing UACL in a binary form (counties with zero UACL versus counties with a non-zero value of UACL) it is the measure of change in rent and land value over time that is important, while a view of only those counties with a non-zero value of UACL reveal that static measures of rent and land value are most important. Regional analyses show that there are some differences in the drivers for counties in one part of the country versus another. However, the observations that measures of rent and land value change over time being most important to whether a county has net UACL or not and that static measures of rent and land value being most important to the amount of UACL in counties that experience a gain hold true for all regions. Third, this dissertation downscales UACL from the county level to the 30m landscape level by using a spatially explicit land change model, Dyna-CLUE. Four counties, each of which is in a distinct agricultural region of the United States, are chosen as study areas. Using rates of increase in UACL from earlier chapters, the model projects UACL emergence over a 50-year period in each county. Patterns and distribution of UACL across each county are measured with landscape metrics to detect whether there are differences among the counties and if some counties are more suitable for renewable energy production. Patterns that are favorable for renewable energy are those that have large, regular patches of UACL. Results show that there are differences in the metrics across the counties and that some emerge as more favorable for renewable energy. Interestingly, two counties that appear similar initially have starkly different patterns and suitability for renewable energy at year 50. This raises the question of whether this difference is due to the initial pattern and characteristics of the county landscape or an artifact of the modeling approach.