Evaluating relative canopy temperature velocity for classifying rice irrigation regimes under simulated calibration drift

Efficient irrigation monitoring is vital for sustainable rice cultivation, especially under alternate wetting and drying (AWD). Thermal canopy-temperature indices are attractive for this purpose; however, conventional absolute-temperature indices—such as canopy temperature depression (CTD) and the crop water stress index (CWSI)—are highly susceptible to the calibration drift inherent in low-cost, uncooled thermal cameras deployed in the field or on UAVs. To address this limitation, the relative canopy temperature velocity (RCTV), a ratio of intra-day canopy-temperature differences, was evaluated against CTD and CWSI for classifying rice irrigation status (sufficient vs. deficit). Two field experiments (2019–2020, two varieties, four irrigation regimes) yielded 294 samples, which were evaluated using seven machine-learning classifiers under date-grouped cross-validation, integrated with SHAP analysis and a 0–5 ℃ calibration-drift simulation under three assumptions about how the wet and dry reference temperatures respond to the drift. Under ideal calibration conditions, RCTV performed comparably to CTD but ranked below CWSI (single-index Macro-F1 scores of 0.66 versus 0.61 and 0.72, respectively). When only the canopy temperature drifted and the references were held fixed, a per-date additive bias of 5 ℃ degraded the Macro-F1 of CWSI and CTD by 28% and 16%, whereas RCTV was unchanged for all seven classifiers because the shared per-date offset cancels in its ratio. so that RCTV overtook CWSI at a drift of 2 ℃. When the wet reference was assumed to drift with the canopy temperature, however, CWSI lost only 12% and remained above RCTV until the drift reached 4 ℃, and when all three temperatures drifted together CWSI was unaffected; independent within-day errors of 1 ℃ reduced RCTV to 0.60. At matched numbers of thermal acquisitions, a single-acquisition CWSI classifier was at least as accurate as the two-acquisition RCTV classifier without drift. The advantage of RCTV is therefore conditional: it is preferable when the calibration error is shared between its acquisition times and the CWSI references are not re-acquired with the drifting camera, at the cost of an additional thermal acquisition. Requiring only canopy and ambient air temperature without external references or absolute calibration, RCTV remains a practical option for uncalibrated thermal sensors for uncalibrated UAV-based rice-irrigation monitoring.

Authors

Institutions

Publication Details

Journal
Agricultural Water Management
Published
2026-11-18
DOI
https://doi.org/10.1016/j.agwat.2026.110956
Primary Topic
Remote Sensing in Agriculture
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Evaluating relative canopy temperature velocity for classifying rice irrigation regimes under simulated calibration drift

Weixing Cao, Yong Tian, Qiang Cao, Yan Zhu et al.
Agricultural Water Management
Remote Sensing in Agriculture
article

Evaluating relative canopy temperature velocity for classifying rice irrigation regimes under simulated calibration drift

Weixing Cao, Yong Tian, Qiang Cao, Yan Zhu, Yinshan Wu, Xiaojun Liu, Zhongwei Li
article en

Abstract

Efficient irrigation monitoring is vital for sustainable rice cultivation, especially under alternate wetting and drying (AWD). Thermal canopy-temperature indices are attractive for this purpose; however, conventional absolute-temperature indices—such as canopy temperature depression (CTD) and the crop water stress index (CWSI)—are highly susceptible to the calibration drift inherent in low-cost, uncooled thermal cameras deployed in the field or on UAVs. To address this limitation, the relative canopy temperature velocity (RCTV), a ratio of intra-day canopy-temperature differences, was evaluated against CTD and CWSI for classifying rice irrigation status (sufficient vs. deficit). Two field experiments (2019–2020, two varieties, four irrigation regimes) yielded 294 samples, which were evaluated using seven machine-learning classifiers under date-grouped cross-validation, integrated with SHAP analysis and a 0–5 ℃ calibration-drift simulation under three assumptions about how the wet and dry reference temperatures respond to the drift. Under ideal calibration conditions, RCTV performed comparably to CTD but ranked below CWSI (single-index Macro-F1 scores of 0.66 versus 0.61 and 0.72, respectively). When only the canopy temperature drifted and the references were held fixed, a per-date additive bias of 5 ℃ degraded the Macro-F1 of CWSI and CTD by 28% and 16%, whereas RCTV was unchanged for all seven classifiers because the shared per-date offset cancels in its ratio. so that RCTV overtook CWSI at a drift of 2 ℃. When the wet reference was assumed to drift with the canopy temperature, however, CWSI lost only 12% and remained above RCTV until the drift reached 4 ℃, and when all three temperatures drifted together CWSI was unaffected; independent within-day errors of 1 ℃ reduced RCTV to 0.60. At matched numbers of thermal acquisitions, a single-acquisition CWSI classifier was at least as accurate as the two-acquisition RCTV classifier without drift. The advantage of RCTV is therefore conditional: it is preferable when the calibration error is shared between its acquisition times and the CWSI references are not re-acquired with the drifting camera, at the cost of an additional thermal acquisition. Requiring only canopy and ambient air temperature without external references or absolute calibration, RCTV remains a practical option for uncalibrated thermal sensors for uncalibrated UAV-based rice-irrigation monitoring.

Agricultural Water ManagementVol. 338
Nanjing Agricultural University (CN), Jiangsu Collaborative Innovation Center for Modern Crop Production (CN)
Zero hunger
Remote Sensing in Agriculture
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon