Spatiotemporal Canopy Temperature Forecasting for Precision Irrigation Management

Abstract

Abstract Canopy temperature (Tc) is a critical physiological indicator of water and heat stress in cotton. Although weather-driven Tc forecasting is used by 60% of Australian cotton growers for irrigation scheduling, current methods typically rely on a single in situ sensor to represent an entire management area. This uniform assumption overlooks substantial spatial variability in Tc and can lead to suboptimal water application. We propose UAV-linear, a novel spatio-temporal forecasting model that integrates high-accuracy in-situ sensors with weekly Unmanned Aerial Vehicle (UAV) thermal imagery to generate high-resolution hourly spatial Tc forecasts. Experimental results show that UAV-linear forecast stress conditions at unmeasured locations as effectively as models trained on exhaustive historical data, achieving a 25-35% improvement over the standard uniform-forecast assumption. Furthermore, in a large-scale validation across 50,000 hectares of commercially active farms (practical dataset), UAV-linear improved stress-hour prediction by 22% relative to the uniform assumption while maintaining accuracy comparable to historical benchmarks. These findings show that the proposed spatio-temporal framework provides the spatial detail needed for differentiated precision management, with potential to improve crop yield and water-use efficiency.

Code and data availability

The supplied blocks describe paper-specific canopy temperature sensor datasets (Carrathool 2020-21, Narrabri 2016, 22 practical farm locations) and UAV thermal imagery, but contain no data availability statement, public repository deposit, or author code/model release. The only public URL mentioned (github.com/Nixtla/…

No evidence-backed public reproduction asset is currently recorded.