Comparison of MODIS Surface Temperatures to In-situ Measurements across the Ablation Zone of the Greenland Ice Sheet
Abstract
The accuracy of land surface temperature (LST) measurements on the Greenland Ice Sheet is critical for successful monitoring of surface processes that affect mass balance. An increasing fraction of Greenland's ice loss is due to surface melt, making summer temperatures especially important, but tracking temperatures year round is needed to observe climate trends and model energy balance processes. Remotely-sensed LST measurements, such as those calculated using data from the Moderate Resolution Imaging Spectroradiometer (MODIS), provide information from the polar regions where in-situ measurements are difficult. This remotely sensed LST data must be validated using ground based measurements in order to ensure accuracy of the temperature calculation algorithms. Using data from the Programme for Monitoring of the Greenland Ice Sheet (PROMICE), we conducted a MODIS LST validation (MOD/MYD11 C6) study over the years 2014 to 2017, using temperature records from seventeen sites across the ablation area of the Greenland Ice Sheet. Our analysis comparing the MODIS and PROMICE datasets is split into two groups: when the in-situ records indicate a surface temperature of zero (called `melt season') and when in-situ temperatures are below zero (called `non-melt season'). Aggregating data from all sites and years in the melt season, discrepancy between datasets is large (RMSE = 6.45°C), but the distribution of discrepancies is centered near zero (Mean Bias: -0.03±6.44°C). In the non-melt season, there is a distinct cold bias in the MODIS LST (Median Bias: 2.19±4.34°C; RMSE = 5.27°C), where MODIS temperatures are lower than in-situ measurements, and this bias is progressively larger as temperatures decrease. To determine what factors covary with the magnitude of the bias, we use a multiple linear regression with solar zenith angle, specific humidity, and LST data from MODIS as predictor variables. In the non-melt season, we find that the MODIS bias is more pronounced at lower temperatures and higher humidity, with these variables explaining anywhere from 10% to 88% (site dependent) of the variability in the discrepancy. MODIS bias is not well predicted by these variables in the melt season. This comparison indicates areas for further exploration of LST algorithms.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMGC127..03Z
- Keywords:
-
- 1616 Climate variability;
- GLOBAL CHANGE;
- 1631 Land/atmosphere interactions;
- GLOBAL CHANGE;
- 1632 Land cover change;
- GLOBAL CHANGE;
- 1640 Remote sensing;
- GLOBAL CHANGE