Disaggregation of remotely sensed land surface temperature: A simple yet flexible index (SIFI) to assess method performances

Disaggregation of land surface temperature (DLST), the aim of which is to generate LSTs with fine resolution, has been attracting increasing attention since the 1980s. The past three decades have been witness to the emergence of DLST methods in large numbers, the accuracies of which were often assessed by comparing the disaggregated with fine spatial resolution LSTs using error indexes such as the root mean square error (RMSE). However, the majority of previous error indexes are, by their nature, insufficient for assessing the performances of DLST methods. This insufficiency is due in part to their lower competence at distinguishing the DLST error from LST retrieval errors and in part to their inability to remove the process controls resulting from different thermal contrasts, temperature units, and resolution ratios among different scenarios in which DLST is conducted. This is also because they are unable to denote the sharpening statuses of the DLST results (e.g., under- or over-sharpening). This status quo has made the evaluation of method performances challenging and sometimes unreliable. To better assess DLST method performances under diversified scenarios, we formulated five protocols, through which a simple yet flexible index (SIFI) was subsequently designed. The establishment of an SIFI includes the following four steps: (1) a detail-based evaluation, which is designed primarily to exclude the impacts of systematic deviations on estimated LSTs; (2) a Gaussian normalization, which is primarily intended to remove the differences in temperature units and thermal contrasts; (3) a triple comparison, with the aim of attenuating the influence of the difference in the resolution ratio in comparisons of method performances; and (4) a piecewise comparison, which is primarily scheduled to distinguish among the three sharpening statuses, under-sharpening, acceptable over-sharpening, and unacceptable over-sharpening. The evaluation ability of SIFI was compared with those of the RMSE, Erreur Relative Globale Adimensionnelle de Synthèse (ERGAS), and image quality index (Q) using simulation tests and actual thermal data. The results illustrate that SIFI generally outperforms the other indexes; it is able to mitigate the impacts from process errors and controls during evaluation and is able to indicate the sharpening statuses accurately. We believe this new index will likely promote the design of future DLST algorithms and procedures.