Thank you for visiting nature.com. You are using a browser version with
limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off
compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site
Table 3: Gap-fill validation outcomes for six validation scenarios, where approximately 5, 10, 20, 30, 40 and 50% of observed (un-filled) remote-sensed MODIS land surface temperature (LST) data were deleted in random spatial clusters to mimic observed patterns of missing observations and subsequently gap-filled.
Quantile regressions, for the median (0.5) quantile, were used to test the relationships between day and night gap−fill prediction errors (absolute difference between observed and predicted temperatures) and support (the number of spatially and temporally neighboring observations, used to estimate the missing value). The mean and minimum number of support cells with observed LST data used to calculate gap-fill predictions for each validation scenario are reported.