Dear friends
I am working on a project in which I have to calculate LST using Landsat imagery. I used different algorithms. However there is a significant difference between LSTs calculated on the imagery and those measured in field. The calculated brightness temperatures are in more accordance with field-measured LSTs. I think that this is because these methods are based on vegetation cover (NDVI) and the study area has a very poor vegetation cover. Is there any other approaches that do not use vegetation indices?
Sincerely Yours
While using the Split-Window (SW) surface temperature retrieval algorithms, in addition to uncertainties in estimating 'Land Surface Emissivity' (LSE) that is derived from 'Fraction of Vegetation Cover' (FVC) which in turn is derived from NDVI, I think you should also take into consideration the Stray Light issue that has plagued Landsat 8 Bands 10 and 11 of the Thermal Infrared Sensor (TIRS) since 2013. As described in the USGS website, 'this stray light increases the reported temperature by up to 4°K in Band 10 and up to 8°K in Band 11.' Although USGS has implemented a correction, the 'results after this correction may still not be considered precise enough for a split-window algorithm application and it is recommended that users refrain from using band 11 data in science studies.' Check out the USGS link for more details.
Although, literature shows that multi-channel algorithms perform better than single channel methods, in case of Landsat 8 TIRS, we are left with no choice but to use single thermal band because of the stray light effect. Therefore, I think, you could try alternative surface temperature algorithms using a single thermal band of Landsat 8 Band 10 or Landsat 7 Band 6 or Landsat 5 Band 6. Some of the algorithms are documented in Qin et al. (2001) and Jimenez-Munoz and Sobrino (2003). Alternatively, you could try with the simplified Radiative Transfer Model (RTM) equation as documented in Barsi et al. (2003).
Regardless of which algorithm you use, estimating LSE is crucial. Most people use NDVI for deriving LSE, however, as you may be aware that this may not be the best method. Therefore, you could try to derive LSE, 1) by Image Classification method which assumes emissivity for each class, or (2) by using emissivity values from spectral libraries such as the ASTER spectral library (http://asterweb.jpl.nasa.gov), or (3) by taking into account the seasonal ASTER Global Emissivity Database (GED) provided by JPL, NASA or (4) In-situ emissivity measurements.
I'm also searching for other non-NDVI based methods to obtain emissivity for calculation of night-time Landsat land surface temperature. If you are aware of any other methods, please let me know.
Hope this helps a bit.
- https://atmcorr.gsfc.nasa.gov/
- https://atmcorr.gsfc.nasa.gov/Barsi_IGARSS03.PDF
- http://scholarworks.rit.edu/theses/9439/
- https://landsat.usgs.gov/how-are-thermal-infrared-sensor-tirs-thermal-bands-aboard-landsat-8-used
A method used by us to calculate LSTs from Landsat images are explained in this paper with references; https://www.researchgate.net/publication/259171924_Remote_sensing_based_analysis_of_urban_heat_islands_with_vegetation_cover_in_Colombo_city_Sri_Lanka_using_Landsat-7_ETM_data
How about applying a bias correction, if the field values show a deviation from the estimated values.
http://www.gis-blog.com/calculation-of-land-surface-temperature-lst-from-landsat-8-using-r/
LSTs from Landsat 5 and 7 images are explained in this paper. Link is below:
https://naldc.nal.usda.gov/download/38199/PDF
It might be nice to try the approach in the literature below for your emissivity computation.
Article Investigating Surface Urban Heat Island Characteristics over...
Hi Vahid,
Why do not you use of MODIS product such as MOD C11? Time resolution is higher relative Landsat.
Hi Vahid,
Please check the link bellow on how to extract LST from Landsat 8 images. Its super easy and fast. The difference between the results and the field measurments was max 2C.
Regards
Article Algorithm for Automated Mapping of Land Surface Temperature ...
hi Vahid
You can use this link: Article Land Surface Emissivity Retrieval from Satellite Data
i recommend using MODIS data as its having high temporal resolution
u can check a few paper here
Article Estimation and validation of Remotely Sensed Land Surface Te...
Article Spatio-temporal variation of land surface temperature and te...
Chapter Estimating Land Surface Temperature and its Lapse Rate over ...
Dear you need to use thermal bands of Landsat 7 or 8 satellites, it can be done from following these steps:
1- convert DN to radiance
2- radiance to at top atmosphere
3- to Reflectance
then use the equation of LST
My regards
You can also read the paper Desertification risk analysis and assessment in Northern Nigeria,
10.1016/j.rsase.2018.04.012
one thing more you need to select a data based on the required accuracy for your project
check out this link
Article Use of normalized difference built-up index in automatically...
What is technique do you use to measure field temperature?
I recommend using a thermal camera that is in contact with the surface
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing