Scholarlink

<p>Soil moisture is the key factor in the hydrological cycle and agricultural productivity, playing a crucial role in understanding climatic variability, drought development, and agricultural risk management. This study presents an integrated framework for multi-regional soil moisture assessment that combines established statistical methods to identify homogeneous spatial groups and detect temporal shifts within them. The proposed framework consists of three phases. In the first phase, Model-Based Clustering (MBC) with the Bayesian Information Criterion (BIC) is applied to produce optimal, homogeneous groups of sampled locations. In the second phase, bootstrapping is used to resample soil moisture observations within each cluster, thereby constructing representative cluster-level time series from homogeneous locations. In the final phase, the Non-Parametric Change Point (NPCP) technique is employed to detect temporal shifts in soil moisture. To validate the proposed framework, we consider the monthly-averaged soil moisture data for 36 districts of Punjab, Pakistan, covering the period from January 1981 to November 2022 (503 months; approximately 42 years). The results of MBC reveal five homogeneous clusters, selected based on the highest BIC value (-10995.2). Bootstrapping was used to construct representative cluster-level soil moisture time series for each identified cluster. NPCP indicated that the most change points occurred in Cluster 2, with 18 shifts, and that the other clusters had between 4 and 6 shifts. The Kolmogorov–Smirnov (KS) test was used as a descriptive measure of distributional differences between adjacent segments following change-point detection. The results demonstrate substantial spatial and temporal variability in soil moisture and highlight the effectiveness of the proposed framework for supporting agricultural planning, irrigation management, and climate adaptation strategies in data-scarce, climate-sensitive regions.</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

Framework for spatiotemporal soil moisture assessment: an application to the integration of model-based clustering with non-parametric change points detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月1日

查看全文
  • Waleeja Tur Rabbia,
  • Asad Ellahi,
  • Mohamed A. E. Khalefa,
  • Walaa A. A. Ismael,
  • Ijaz Hussain,
  • Paulo Canas Rodrigues

摘要

Soil moisture is the key factor in the hydrological cycle and agricultural productivity, playing a crucial role in understanding climatic variability, drought development, and agricultural risk management. This study presents an integrated framework for multi-regional soil moisture assessment that combines established statistical methods to identify homogeneous spatial groups and detect temporal shifts within them. The proposed framework consists of three phases. In the first phase, Model-Based Clustering (MBC) with the Bayesian Information Criterion (BIC) is applied to produce optimal, homogeneous groups of sampled locations. In the second phase, bootstrapping is used to resample soil moisture observations within each cluster, thereby constructing representative cluster-level time series from homogeneous locations. In the final phase, the Non-Parametric Change Point (NPCP) technique is employed to detect temporal shifts in soil moisture. To validate the proposed framework, we consider the monthly-averaged soil moisture data for 36 districts of Punjab, Pakistan, covering the period from January 1981 to November 2022 (503 months; approximately 42 years). The results of MBC reveal five homogeneous clusters, selected based on the highest BIC value (-10995.2). Bootstrapping was used to construct representative cluster-level soil moisture time series for each identified cluster. NPCP indicated that the most change points occurred in Cluster 2, with 18 shifts, and that the other clusters had between 4 and 6 shifts. The Kolmogorov–Smirnov (KS) test was used as a descriptive measure of distributional differences between adjacent segments following change-point detection. The results demonstrate substantial spatial and temporal variability in soil moisture and highlight the effectiveness of the proposed framework for supporting agricultural planning, irrigation management, and climate adaptation strategies in data-scarce, climate-sensitive regions.