Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4531574 | Continental Shelf Research | 2016 | 10 Pages |
•Secchi Disk Depth (SDD) values were considered as an indicator for clarity.•Landsat 8 OLI bands were nominated for estimating SDD values.•An overall model was proposed to estimate SDD values from Landsat bands.•Optimum models were statistically examined to find the best one.•Best model was calibrated for the study area.
This study examined the advantages of incorporating the new band of Landsat-8 OLI imagery (band 1: Coastal/Aerosol, 435–451 nm) to a model for estimation of Secchi disk depth (SDD) values (as an indicator for transparency) in near-shore coastal waters using multispectral bands. In doing so, Chabahar Bay in the southern part of Iran (north of Gulf of Oman) was selected as the study area. Two approximately four-hour in-situ observations (including 48 and 56 field measured SDD values for each date respectively) were performed in the study area using Secchi disk; this was designed to start about two hours before and end about two hours after the time of satellite overpasses. Thereafter, a model was formed for estimation of SDD values based on the terms including all possible linear and mutual ratio values of Coastal/Aerosol (B1), Blue (B2), Green (B3), and Red bands (B4). In the first step, the correlation between reflectance/ratio reflectance values of these bands and Ln(SDD) values were calculated to indicate higher correlated bands/band ratios with the first field measured SDD values. Consequently, 17 combinations of highest correlated bands/band ratios were selected to estimate SDD values. In this regard, 32 points among the 48 field observations were selected to determine unknown coefficients of models using a multiple linear regression, and the rest 16 points were designated for accuracy assessment the results. Eventually, the measured SDD values in second field observations were utilized for validating the results. Final results demonstrated that combination of linear terms including B1, B2 and B3 bands and band ratio terms including ratio reflectance values of B4/B3, B3/B1, and B2/B1 has led to obtain the highest accuracy (R2=0.866 and RMSE=0.919, SVM feature weight=4.294). This was in agreement with the results obtained from the second observations. Finally, by applying the entire 104 field observed SDD values, the model in form of SDD=0.077exp(1.209RB1−1.739RB2+0.412RB3−5.198RB3/RB1+10.408RB2/RB1) can be used to determine SDD values in near-shore coastal waters of Chabahar Bay using atmospherically corrected reflectance values of Landsat-8 OLI bands.