Scholarlink

<p>Accurate estimation of the Normalized Difference Vegetation Index (NDVI) is essential for agricultural monitoring, hydrological studies, and environmental assessments. However, remotely sensed NDVI data typically have low temporal resolution due to atmospheric interference, cloud cover, and sensor limitations, which pose challenges for practical data utilization. In this study, we applied machine learning algorithms, specifically Random Forest (RF) and Artificial Neural Network (ANN), to predict NDVI using meteorological variables, topographic features, and drought indices. The study area was set in Gangneung, Donghae, and Samcheok, which are representative forest regions in South Korea. We constructed a 1&#xa0;km resolution grid-based dataset for model training and validation. The performance of the RF and ANN models was evaluated using correlation coefficient (CC), root mean square error (RMSE), percent bias (PBIAS), and Nash–Sutcliffe efficiency (NSE). The results showed that the RF model achieved higher accuracy in NDVI prediction. In the Donghae region, the RF model yielded an RMSE of 655.6, CC of 0.95, NSE of 0.91, and PBIAS of 0.05%, while the ANN model resulted in an RMSE of 1099.4, CC of 0.86, NSE of 0.73, and PBIAS of −&#xa0;0.02%, demonstrating overall superior performance. Additionally, the models captured the seasonal variability of NDVI. This study highlights the potential of machine learning for improving the reliability of NDVI data, contributing to enhanced accuracy in ecological and agricultural research.</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

Temporal resolution enhancement of NDVI time series using machine learning approaches

  • 期刊论文

  • 发表日期：

    2025年7月26日

查看全文
  • Soonchan Park,
  • Changju Kim,
  • Heechan Han

摘要

Accurate estimation of the Normalized Difference Vegetation Index (NDVI) is essential for agricultural monitoring, hydrological studies, and environmental assessments. However, remotely sensed NDVI data typically have low temporal resolution due to atmospheric interference, cloud cover, and sensor limitations, which pose challenges for practical data utilization. In this study, we applied machine learning algorithms, specifically Random Forest (RF) and Artificial Neural Network (ANN), to predict NDVI using meteorological variables, topographic features, and drought indices. The study area was set in Gangneung, Donghae, and Samcheok, which are representative forest regions in South Korea. We constructed a 1 km resolution grid-based dataset for model training and validation. The performance of the RF and ANN models was evaluated using correlation coefficient (CC), root mean square error (RMSE), percent bias (PBIAS), and Nash–Sutcliffe efficiency (NSE). The results showed that the RF model achieved higher accuracy in NDVI prediction. In the Donghae region, the RF model yielded an RMSE of 655.6, CC of 0.95, NSE of 0.91, and PBIAS of 0.05%, while the ANN model resulted in an RMSE of 1099.4, CC of 0.86, NSE of 0.73, and PBIAS of − 0.02%, demonstrating overall superior performance. Additionally, the models captured the seasonal variability of NDVI. This study highlights the potential of machine learning for improving the reliability of NDVI data, contributing to enhanced accuracy in ecological and agricultural research.