Remote Sensing of Chlorophyll‐A in Case II Waters: A Novel Approach With Improved Accuracy Over Widely Implemented Turbid Water Indices

Abstract

A new semianalytical algorithm was formulated to retrieve chlorophyll‐a (CHL) in optically complex waters using in situ data set of coastal waters of eastern Arabian Sea. The algorithm was derived using CHL index of the form, x = (Rrs(λ1)−1−Rrs(λ2)−1) × Rrs(λ3). The first wavelength (λ1) represents the secondary peak of CHL, while the second wavelength (λ2) and third wavelength (λ3) were delineated using a radiative transfer model and partial derivative analysis of hyperspectral remote sensing reflectance, respectively. Further iteration of three wavelengths between 600 and 700 nm resulted in a two‐wavelength index, x = (Rrs(λ1)−1−Rrs(λ2)−1) × Rrs(λ2). This was further regressed with CHL data initially used for three wavelength index. The final form of algorithm, Goa University Case II (GUC2), cMCHL=113.112x3−58.408x2+8.669x − 0.0384, was validated with in situ CHL ranging between 0.11 and 25.56 μg/L, resulted in a strong correlation r2 = 0.99, RMSE = 0.30, and bias = 0.03. A comparison with NIR‐Red two‐band, three‐band, four‐band, synthetic chlorophyll index, and normalized difference chlorophyll index pointed to the nonsuitability of turbid water indices in different water types of the study area. For the first time, a CHL algorithm has been tested successfully in water types outside the region of its formulation. A pixel‐to‐pixel validation of GUC2‐derived MERIS CHL with NASA bio‐Optical Marine Algorithm Dataset and Satellite Coastal and Oceanography Research data set resulted in correlation, bias, and RMSE of 0.90, −0.0013, and 1.2499, respectively. Furthermore, GUC2 was successfully tested in Chesapeake Bay for accurate retrieval of CHL from stations with varying turbidity levels.

Studying Optical Characteristics of Diffused Light Reflecting from Naturally Senescing Leaves of Deciduous Trees

“…In the range 350-500 nm, all the leaf samples under study are manifesting some low reflectance coefficient (RC) of approximately 5 %. This is not surprising, since the chlorophyll-a and chlorophyll-b absorption peaks are respectively appearing at wavelengths of 440 and 460 nm [9,11,12,[19][20][21][22]. Rapid increase in reflectivity of the leaf samples under study is observed in the range of 490-520 nm values [11].…”

Characteristics of Mudbanks along the Southwest Coast of India

“…In addition, a bimodal variability of chlorophyll-a concentration during summer monsoon occurs in this region. A new algorithm, Goa University Case II (GUC2), was developed to retrieve chl-a in optically complex waters and its validation with in situ observations from the eastern coastal Arabian Sea is useful for the chlorophyll estimations from this region [79]. GUC2 could identify two appropriate wavelengths for deriving chl-a and it could eliminate the effects of total suspended matter and coloured dissolved organic matter.…”

WaterNet: A Convolutional Neural Network for Chlorophyll-a Concentration Retrieval

“…Band expansion phase. In line with the works of [5,49,50], who utilized band combinations as feature candidates in Chl-a concentration modeling, this phase aims to enrich the spectral information using band combination. The filter's kernel size used in this phase is 1 × 1 × 3, indicating that the convolution is performed in the spectral domain and that spectral enrichment is achieved using linear band combination.…”

“…To obtain fair comparisons, we set the numbers of unknown parameters in WaterNet and the feedforward neural networks to be almost the same. The number of neurons in the input, hidden, and output layers of the three feedforward networks are (16,264,1), (16,70,50,1), and (16,44,44,44,1), respectively; whereas the numbers of unknowns are 4753, 4791, and 4753, respectively, as shown in Table 6. The proposed two-stage training is applied to WaterNet and the feedforward neural networks for a fair comparison.…”