Abstract
Land Surface Temperature (LST) as provided by remote sensing onboard satellites is a key parameter for a number of applications in Earth System studies, such as numerical modelling or regional estimation of surface energy and water fluxes. In the case of Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Terra or Aqua, pixels have resolutions near 1 km 2 , LST values being an average of the real subpixel variability of LST, which can be significant for heterogeneous terrain. Here, we use Landsat 7 LST decametre-scale fields to evaluate the temporal and spatial variability at the kilometre scale and compare the resulting average values to those provided by MODIS for the same observation time, for the very heterogeneous Campus of the University of the Balearic Islands (Mallorca, Western Mediterranean), with an area of about 1 km 2 , for a period between 2014 and 2016. Variations of LST between 10 and 20 K are often found at the sub-kilometre scale. In addition, MODIS values are compared to the ground truth for one point in the Campus, as obtained from a four-component net radiometer, and a bias of 3.2 K was found in addition to a Root Mean Square Error (RMSE) of 4.2 K. An indication of a more elaborated local measurement strategy in the Campus is given, using an array of radiometers distributed in the area.
Highlights
Land Surface Temperature (LST) is the radiative temperature of the most superficial part of the soil and vegetation of an element of the surface, sometimes referred to as the skin temperature of the surface [1]
An example of LST difference between ETM+ and Moderate Resolution Imaging Spectroradiometer (MODIS) products is shown in Figure 2, for a scene from 8 November 2015 at 10:30 UTC
The UIB Campus is defined with a white square that consists of 33 × 33 Landsat 7 ETM+ pixels at 30 m spatial resolution (Figure 2c, corresponding to the disaggregated Thermal infrared (TIR) scene from the original 60 m spatial resolution)
Summary
Land Surface Temperature (LST) is the radiative temperature of the most superficial part of the soil and vegetation of an element of the surface, sometimes referred to as the skin temperature of the surface [1]. All these validation works have demonstrated good operational performance of the three MODIS LST algorithms at different surface types with discrepancies respect to reference data of around 1 K. To use LST values given by a sensor onboard a satellite, these must have gone through processes of validation and calibration that provide an estimation of the uncertainty of the value This information is obtained primarily with ground-based data used for comparison, usually for ground homogeneous conditions. It may well happen that the pixel-averaged value or the local measurement point are not representative of the actual conditions governing the surface atmosphere exchanges over each of the subpixel tiles This would be important since it could provide inadequate values of surface temperature or sensible and latent energy fluxes for numerical models or agricultural applications [24]. Landsat 7 pixel, allowing assess to the possible discrepancies made in the validation results when taking these three quantities as equivalent
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
