Methodology
After reviewing literature on turbidity, it was clear that three turbidity algorithms are the standard for turbidity models: Nechad, Nechad 2016 and Dogliotti.
These turbidity algorithms rely on the red band of surface reflectance. We chose to use the Nechad 2016 algorithm. We have been quantifying differences in turbidity generated from these products:
- Landsat 8 Surface Reflectance product in GEE
- Sentinel 2 Surface Reflectance product in GEE
- ACOLITE Surface Reflectance
2016 Nechad algorithm is as follows and is performed at each pixel:
T = turbidity, A_t = coefficient, P_w = water leaving reflectance, C = coefficient
Our group utilized surface reflectance as opposed to water leaving reflectance. The coefficients for the turbidity were obtained from the source code in ACOLITE which is a software program used to atmospherically-correct images for coastal or inland water applications. We validated our implementation of the turbidity algorithm with ACOLITE’s and found no differences at the pixel level.
Through this project we aimed to quantify what the differences were in both surface reflectance and derived turbidity from ACOLITE and Google Earth Engine’s engine pre-processed images. We also compare these turbidity products to in-situ measurements at both locations.
Our workflow required running ACOLITE’s software on multiple Landsat 8 and Sentinel 2 scenes that we downloaded. The output from ACOLITE (surf_refl and turbidity) were uploaded to GEE to compare with GEE’s surf_refl and turbidity for both satellites.