Ndlovu et al. (2025) Enhancing the estimation of equivalent water thickness in neglected and underutilized taro crops using UAV acquired multispectral thermal image data and index-based image segmentation

Identification

• Journal: Remote Sensing Applications Society and Environment
• Year: 2025
• Date: 2025-10-25
• Authors: Helen S. Ndlovu, John Odindi, Mbulisi Sibanda, Onisimo Mutanga
• DOI: 10.1016/j.rsase.2025.101758

Research Groups

• Discipline of Geography and Environmental Science, School of Agricultural Earth and Environmental Sciences, University of KwaZulu-Natal, Scottsville, Pietermaritzburg, South Africa
• Department of Geography, Environmental Studies & Tourism, Faculty of Arts, University of the Western Cape, Bellville, South Africa

Short Summary

This study evaluated the use of UAV-acquired multispectral and thermal imagery, combined with index-based segmentation, to estimate the equivalent water thickness (EWT) of taro crops in smallholder farmlands. The findings demonstrate that incorporating thermal data significantly improves EWT prediction accuracy, with the Excess Green minus Excess Red (ExGR) technique proving highly effective.

Objective

• To evaluate the applicability of multispectral and thermal infrared UAV imagery in understanding taro’s water status by predicting its canopy equivalent water thickness (EWT) in smallholder farmlands.

Study Configuration

• Spatial Scale: Local, farm-scale monitoring of smallholder farmlands.
• Temporal Scale: Not explicitly defined, but implies short-term, specific acquisition periods for "near-real-time" monitoring.

Methodology and Data

• Models used: Deep learning techniques, three index-based segmentation techniques (specifically Excess Green minus Excess Red (ExGR)).
• Data sources: Unmanned aerial vehicles (UAVs) equipped with high-resolution multispectral and thermal infrared imagery.

Main Results

• A significant difference (P < 0.05) was observed in the prediction accuracies of taro EWT when including the thermal band compared to excluding it.
• When including the thermal band, the prediction achieved an R² of 0.92, a Root Mean Square Error (RMSE) of 8.04 g/m², and a relative Root Mean Square Error (rRMSE) of 15.31 %.
• Excluding the thermal band resulted in an R² of 0.91, an RMSE of 8.73 g/m², and an rRMSE of 16.64 %.
• The Excess Green minus Excess Red (ExGR) technique was identified as valuable for accurately predicting canopy EWT.

Contributions

• Provides a foundational framework for monitoring the water status of neglected and underutilized crops like taro using UAV-acquired multispectral and thermal imagery.
• Highlights the significant improvement in equivalent water thickness (EWT) prediction accuracy by incorporating thermal remote sensing data.
• Demonstrates the effectiveness of index-based image segmentation, particularly the ExGR technique, for EWT estimation in taro.

Funding

• Not specified in the provided text.

Citation

@article{Ndlovu2025Enhancing,
  author = {Ndlovu, Helen S. and Odindi, John and Sibanda, Mbulisi and Mutanga, Onisimo},
  title = {Enhancing the estimation of equivalent water thickness in neglected and underutilized taro crops using UAV acquired multispectral thermal image data and index-based image segmentation},
  journal = {Remote Sensing Applications Society and Environment},
  year = {2025},
  doi = {10.1016/j.rsase.2025.101758},
  url = {https://doi.org/10.1016/j.rsase.2025.101758}
}