Scholarlink

Abstract <p>This study examines the spatiotemporal dynamics of groundwater levels in the Gorgan Plain, northern Iran, utilizing 16 years (2007–2022) of monthly observations from 29 semi-deep wells. The depth of the selected shallow wells ranged from 35 to 60 m. Space-Time Autoregressive Moving Average (STARMA) models, incorporating first-order (STARMA-1) and second-order (STARMA-2) neighborhood structures, were developed to forecast shallow groundwater levels over a 12-month horizon and were benchmarked against the temporal Seasonal Autoregressive Integrated Moving Average (SARIMA) model. Model performance, evaluated using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), demonstrated that both STARMA models consistently outperformed SARIMA across all wells. STARMA-2 yielded slightly higher predictive accuracy than STARMA-1 for most wells, although their performance differences were minor. The SARIMA model, although less precise (average RMSE = 33.71 cm), demonstrated a reasonable capability in capturing groundwater trends compared to STARMA-1 (RMSE = 9.41 cm) and STARMA-2 (RMSE = 5.50 cm). These findings underscore the importance of accounting for spatial dependencies in groundwater forecasting, particularly in data-rich semi-arid aquifers. The results align with global evidence from GRACE satellite missions and USGS monitoring programs, highlighting the critical role of spatiotemporal frameworks in sustainable groundwater management under increasing climatic and anthropogenic conditions.</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

Modeling of Groundwater Level Dynamics in Gorgan Plain, Iran: a Spatiotemporal ARMA Approach

  • 期刊论文

  • 发表日期：

    2026年5月4日

查看全文
  • Mohammad Khatiripour,
  • Nader Jandaghi,
  • Mojtaba Ghareh Mahmoodlu,
  • Majid Azimmohseni

摘要

Abstract

This study examines the spatiotemporal dynamics of groundwater levels in the Gorgan Plain, northern Iran, utilizing 16 years (2007–2022) of monthly observations from 29 semi-deep wells. The depth of the selected shallow wells ranged from 35 to 60 m. Space-Time Autoregressive Moving Average (STARMA) models, incorporating first-order (STARMA-1) and second-order (STARMA-2) neighborhood structures, were developed to forecast shallow groundwater levels over a 12-month horizon and were benchmarked against the temporal Seasonal Autoregressive Integrated Moving Average (SARIMA) model. Model performance, evaluated using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), demonstrated that both STARMA models consistently outperformed SARIMA across all wells. STARMA-2 yielded slightly higher predictive accuracy than STARMA-1 for most wells, although their performance differences were minor. The SARIMA model, although less precise (average RMSE = 33.71 cm), demonstrated a reasonable capability in capturing groundwater trends compared to STARMA-1 (RMSE = 9.41 cm) and STARMA-2 (RMSE = 5.50 cm). These findings underscore the importance of accounting for spatial dependencies in groundwater forecasting, particularly in data-rich semi-arid aquifers. The results align with global evidence from GRACE satellite missions and USGS monitoring programs, highlighting the critical role of spatiotemporal frameworks in sustainable groundwater management under increasing climatic and anthropogenic conditions.