Scholarlink

<p>Long-term land use projections are essential for food security planning, conservation policy, and sustainable development, yet forecasting frameworks applied to the complete historical record of United States land use remain absent from the literature. Here we present the first comprehensive forecasting analysis of the USDA Economic Research Service (ERS) Major Land Uses (MLU) dataset, which spans 1945 to 2017 across 16 temporal observations for the 48 contiguous states. We develop and compare three forecasting approaches: (1) a Markov chain transition probability model estimated from 15 consecutive period-pairs via constrained least squares; (2) Akaike Information Criterion (AIC)-selected parametric curve fitting among linear, quadratic, logistic, and exponential models; and (3) scenario-modified Markov projections representing business-as-usual (BAU), accelerated urbanization, and conservation pathways. Projections extend 50 years to 2067 with bootstrap-derived uncertainty bounds from 500 iterations. Under BAU, cropland is projected to decline from 20.6% to 17.1% of total land area, forest-use land from 28.0% to 19.3%, while grassland pasture and range increases from 34.8% to 39.8% and special uses from 9.0% to 12.6%. AIC model selection independently identifies logistic saturation curves as the best-fitting model for cropland, forest, and urban land, providing convergent evidence that these major transitions are approaching asymptotic equilibria rather than continuing linearly. Under accelerated urbanization, urban land reaches 6.5% by 2067 with correspondingly greater losses in cropland and forest. State-level Markov models reveal convergence half-lives ranging from 5 to 15 years, demonstrating that the U.S. land use transformation is geographically asynchronous. The proposed framework leverages publicly available census-based tabular data and is designed to complement, rather than replace, remote sensing approaches; it can be readily adapted to any country that maintains periodic census-based or survey-based land use inventories, particularly in contexts where high-frequency national accounting over multi-decadal periods is required.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting US land use through 2067 using multi-method projections applied to seven decades of USDA data

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月16日

查看全文
  • Nasrin Alamdari

摘要

Long-term land use projections are essential for food security planning, conservation policy, and sustainable development, yet forecasting frameworks applied to the complete historical record of United States land use remain absent from the literature. Here we present the first comprehensive forecasting analysis of the USDA Economic Research Service (ERS) Major Land Uses (MLU) dataset, which spans 1945 to 2017 across 16 temporal observations for the 48 contiguous states. We develop and compare three forecasting approaches: (1) a Markov chain transition probability model estimated from 15 consecutive period-pairs via constrained least squares; (2) Akaike Information Criterion (AIC)-selected parametric curve fitting among linear, quadratic, logistic, and exponential models; and (3) scenario-modified Markov projections representing business-as-usual (BAU), accelerated urbanization, and conservation pathways. Projections extend 50 years to 2067 with bootstrap-derived uncertainty bounds from 500 iterations. Under BAU, cropland is projected to decline from 20.6% to 17.1% of total land area, forest-use land from 28.0% to 19.3%, while grassland pasture and range increases from 34.8% to 39.8% and special uses from 9.0% to 12.6%. AIC model selection independently identifies logistic saturation curves as the best-fitting model for cropland, forest, and urban land, providing convergent evidence that these major transitions are approaching asymptotic equilibria rather than continuing linearly. Under accelerated urbanization, urban land reaches 6.5% by 2067 with correspondingly greater losses in cropland and forest. State-level Markov models reveal convergence half-lives ranging from 5 to 15 years, demonstrating that the U.S. land use transformation is geographically asynchronous. The proposed framework leverages publicly available census-based tabular data and is designed to complement, rather than replace, remote sensing approaches; it can be readily adapted to any country that maintains periodic census-based or survey-based land use inventories, particularly in contexts where high-frequency national accounting over multi-decadal periods is required.