EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

An unbiased spatiotemporal fusion approach to generate daily 100 m spatial resolution land surface temperature over a continental scale

Yi Yu1,2, Luigi Renzullo1, Siyuan Tian1, and Brendan Malone2
Yi Yu et al.
  • 1Fenner School of Environment & Society, The Australian National University, Canberra, ACT 2601, Australia
  • 2CSIRO Agriculture and Food, Canberra, ACT 2601, Australia

High spatial resolution land surface temperature (LST) (<= 100 m) has a considerable significance for small scale studies like agricultural applications and urban heat island studies. Originally developed for optical data, spatiotemporal fusion methods, such as the widely used Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) and the Enhanced STARFM (ESTARFM), are gradually becoming promising approaches to generate high resolution thermal variables but still have shortcomings, such as an invalid assumption in thermal fields and the accumulation of systematic biases. Hence, we proposed a variant of the ESTARFM algorithm, referred as the unbiased ESTARFM (ubESTARFM), aiming to better accommodate the spatiotemporal approach to thermal studies. We evaluated the results derived from our method and the typical ESTARFM against both in-situ LST and the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) LST over a continental scale of Australia. The results show that the ubESTARFM has a bias of 2.55 K, unbiased RMSE (ubRMSE) of 2.57 K, and Pearson correlation coefficient (R) of 0.95 against the in-situ LST over 11290 samples at 12 sites, all of which are significantly better than that of the ESTARFM, with a bias of 4.73 K, ubRMSE of 3.80 K and R of 0.92. In the cross-satellite comparison, the ubESTARFM LST has a bias of -1.69 K, ubRMSE of 2.00 K, and R of 0.70 over 43 near clear-sky scenes, while the ESTARFM LST has a bias of 1.79 K, ubRMSE of 2.68 K, and R of 0.59. Overall, the ubESTARFM is able to avoid the accumulation of systematic bias, considerably reduce the deviation of uncertainty, and maintain a good level of correlation with validation datasets compared to the typical ESTARFM algorithm. It is a promising method to integrate reliable numeric values from coarse resolution LST and spatial heterogeneity from fine resolution LST, and may be further coupled with energy balance or radiative transfer models to better enable farm- or regional-scale water management strategy or decision making.

How to cite: Yu, Y., Renzullo, L., Tian, S., and Malone, B.: An unbiased spatiotemporal fusion approach to generate daily 100 m spatial resolution land surface temperature over a continental scale, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-1501,, 2023.