Testing the accuracy of solar‐reflected radiation corrections applied during satellite shortwave infrared thermal analysis of active volcanoes

Abstract

Shortwave infrared (SWIR) satellite imagery is frequently used to study thermal emission from active volcanoes. When using daytime SWIR observations it is necessary to first isolate the thermally emitted signal component, principally by subtracting contributions due to solar‐reflected radiation. Two differing approaches have been used: (1) A “mean” approach based on the average SWIR reflectance of the nonthermally anomalous volcanic background, and (2) a “per‐pixel” approach which estimates the SWIR reflectance of each pixel using a measure of their near infrared reflectance. We assess the accuracy of these using near‐coincident daytime and nighttime Landsat Thematic Mapper (TM) imagery of the active lava dome of Unzen Volcano. Using the nighttime data as a reference we find the most commonly applied mean daytime correction approach to be seriously inadequate, returning results for thermally radiant area 81–94% smaller than the nighttime values. The per‐pixel correction method appears somewhat more effective but still provides values 47–61% smaller than the reference data set. Daytime underestimation of total SWIR thermal spectral radiance is less severe than that of total thermally radiant area because it is primarily pixels having low thermal signals that pose a problem for the daytime approaches. Total thermal spectral radiance is underestimated by 52–79% using the daytime data with the mean correction approach, and by 28–59% with the per‐pixel approach. When applying the TM “dual‐band” technique to determine subpixel hot spot temperatures from the SWIR thermal signals, we find that surfaces cooler than 400°C are poorly represented in the daytime retrievals. However, when using the per‐pixel daytime data correction method the distribution of retrieved hot spots >400°C corresponds quite well to the pattern derived via nighttime imagery. Due to the significant parameter underestimation found to be inherent when using daytime data, this study further indicates the value of nighttime observations to volcanic SWIR thermal studies. If daytime data are used, then we recommend use of the per‐pixel solar‐reflected radiation correction method but would advise caution when quantitatively interpreting the derived thermal parameters, at least at andesitic and dacitic volcanoes. It is possible that daytime retrievals may be more robust at basaltic volcanoes due to their characteristic higher magmatic temperatures and lower surface reflectances.