Advancing spatiotemporal monitoring capability of agricultural drought through integrated use of land surface temperature and soil moisture

Abstract

Agricultural drought is a phenomenon that occurs when there is insufficient soil moisture (SM) to support crop growth and ecosystem functioning, accompanied by land surface temperature (LST) anomalies indicating reduced evaporative cooling. Understanding its spatiotemporal dynamics and mechanism is important in water-limited environments like Australia. With the emerging capabilities of remote sensing (RS) data, it remains underexplored how effectively RS data can monitor the spatiotemporal variability of agricultural drought at the field scale. Specifically, three key limitations were identified: i) the trade-off between spatial and temporal resolution in satellite observations, and inconsistencies between different platforms; ii) the spatial mismatch between point-scale measurements and gridded RS data; and iii) limited understanding of how drought mechanisms operate across different spatial scales.

This thesis addressed these challenges through four interconnected studies. To overcome the spatial and temporal limitations of LST data, we first developed an unbiased spatiotemporal fusion approach, specifically designed to accommodate the high temporal dynamics of LST data. Concurrently, we addressed the inter-platform inconsistency limitation through a solar geometry-based approach that harmonises geostationary LST retrievals with data from well-established polar-orbiting platforms, which quantifies and corrects for the spatial heterogeneity of diurnal LST discrepancies. These two studies can jointly enable the estimation of sub-daily LST at field scales, essential for capturing rapid fluctuations of drought dynamics in heterogeneous agricultural landscapes.

The third study addressed the spatial mismatch between point-scale measurements and satellite footprints, to generate field-scale SM estimates that maintain physical consistency with concurrent LST dynamics. We implemented a machine learning (ML) framework that upscaled in-situ SM to field scale, while quantifying prediction uncertainties through a metric named the Area of Applicability (AOA). This approach delineates the spatial domain where predictions maintain statistical reliability at the satellite footprint, enhancing our capacity to interpret water stress dynamics across agricultural landscapes with varying degrees of confidence.

The fourth study integrated the developed methods and products, to enhance agricultural drought characterisation and mechanistic understanding across spatiotemporal scales. We developed a downscaled Soil Moisture Agricultural Drought Index (dSMADI), which integrates SM, thermal stress, and lagged vegetation response information. The dSMADI demonstrated capability to capture field-scale agricultural water stress patterns during the "Tinderbox" drought in 2019, revealing spatial heterogeneity that was obscured in coarser resolution products. Scale-dependent analysis revealed shifting contributions of individual components across spatial scales. Specifically, thermal stress and lagged vegetation response co-dominated the drought variations in agricultural regions at the continental scale, while moisture dynamics exhibited a mediating influence in agricultural water stress manifestation at both field scales and flux tower footprints.

Overall, this thesis contributed to advancing the capability of RS data, including LST and SM, for monitoring and understanding agricultural drought across multiple spatiotemporal scales. By addressing key methodological limitations through integrated approaches, we presented approaches that enhance the practical utility of RS data for characterising drought dynamics. The complementary nature of the developed techniques, from spatiotemporal fusion to uncertainty quantification, provides a pathway toward more reliable field-scale drought assessment in water-limited environments, and ultimately contributes to improving drought resilience in agricultural systems. Show more