Reliability of RESOLV™ Software Across Very Different Environments
The RESOLV™ service is built around the new technology of CMAC. We evaluated CMAC against the most popular software systems for atmospheric correction of satellite data: for Sentinel 2’s Sen2Cor and for Landsat’s LaSRC processing. This testing applied robust surrogate estimates of surface reflectance, analysis of error and the qualitative appearance of corrected images. Of these two, LaSRC performed more accurately than Sen2Cor. LaSRC is also the mathematical basis for a harmonization pathway proposed for the conversion smallsat data to surface reflectance. The analyses were mostly confined to warehouse/industrial district areas of interest (AOIs) with high spectral diversity: CMAC and LaSRC both performed well and agreed closely.
We asked the question, how well do CMAC and LaSRC perform when applied to a very different natural environment that has relatively low spectral diversity? We answered this question using a null hypothesis based on the fact that CMAC and LaSRC agreed closely for analyses of AOIs of high spectral diversity. The null hypothesis states:
Equivalent top-of-atmosphere reflectance (TOAR) values from images of different environments collected under the same level of atmospheric effect, when atmospherically corrected, will yield equivalent surface reflectance (i.e., no difference). Translated to simpler terms: “the same input, affected by the same conditions will yield the same output.”
The test area to examine this question was an area of low spectral diversity, shortgrass prairie in Alberta, Canada, whose surface reflectance was expected to remain constant during the 16-day test period between July 29 and August 14, 2023. The results were definitive. LaSRC results showed serious divergence between the two environments locations of low and high spectral diversity: only 49.7% (blue band) to 97.2% (NIR band) agreement between the two locations: null hypothesis rejected. This result should be of especial interest for agriculture because, like shortgrass prairie, cultivated fields at virtually all stages can exhibit low spectral diversity.
CMAC reflectance remained consistent between the nearly disparate environments. With 99.4% (blue band) to 99.9% (NIR) agreement between the results from two locations, the null hypothesis was accepted. These results conclusively demonstrated CMAC surface reflectance estimates were robust and reliable.
Landsat 8 views of Lake Diefenbaker, Saskatchewan, CA, June 15, 2020 before and after RESOLV.
About the Author
Hello,
I’m Dr. David Groeneveld, founder and leader of RESOLV™. Our software atmospherically corrects smallsat data conveniently, accurately and reliably and does so in near real time. The benefits of RESOLV™ go beyond its technical capabilities. Better accuracy helps researchers, scientists, and others make smarter choices to monitor and manage our planet.
Curious to learn more about RESOLV™, the science behind it and its potential for correcting smallsat images? Fill out this short form and I’ll be in touch.