Remote sensing inversion study of total organic carbon concentration in Karst Plateau Lakes–Taking Pingzhai reservoir as an example

Figures & data

Figure 1. Location of the Pingzhai Reservoir.

Table 1. Parameters for each band of the Sentinel-2.

Bands	Bands name	Central wavelength/nm	Wave width	Resolution/m
B1	Coastal aerosol	442.3	45	60
B2	Blue	492.1	98	10
B3	Green	559	46	10
B4	Red	665	39	10
B5	Vegetation Red Edege	703.8	20	20
B6	Vegetation Red Edege	739.1	18	20
B7	Vegetation Red Edege	779.7	28	20
B8	NIR	833	133	10
B8a	Vegetation Red Edege	864	32	20
B9	Water vapour	943.2	27	60
B10	SWIR-Cirrus	1376.9	76	60
B11	SWIR	1610.4	141	20
B12	SWIR	2185.7	238	20

Figure 2. BP neural network building process.

Figure 3. Pearson correlation coefficients of sentinle-2 1-8 bands with TOC concentrations.

Table 2. Pearson’s correlation coefficients of single band and band combinations with TOC concentrations.

Bands	Pearson	Bands	Pearson	Bands	Pearson	Bands	Pearson
b1	−0.3011	b6 + b7	−0.6569	(b6-b7)/(b6 + b7)	0.3374	(b6-b7)/b8	0.4875
b2	−0.2681	b6 + b8	−0.6424	(b6 + b7)/(b6-b7)	0.1534	(b6-b8)/b7	0.0954
b3	−0.4135	b7 + b8	−0.6558	(b6-b8)/(b6 + b8)	−0.0074	(b7-b8)/b6	−0.1668
b4	−0.3914	b6*b7	−0.6599	(b6-b8)/(b6 + b8)	−0.3548	b8/(b6 + b7)	0.0429
b5	−0.5717	b6*b8	−0.6342	(b7-b8)/(b7 + b8)	−0.1537	b7/(b6 + b8)	−0.3183
b6	−0.6425	b7*b8	−0.6752	(b7 + b8)/(b7-b8)	0.2424	b6/(b7 + b8)	0.2823
b7	−0.667	b6/b7	0.3331	(b6 + b7)/b8	−0.953	b8/(b6-b7)	0.1739
b8	−0.6283	b6/b8	−0.0348	(b6 + b8)/b7	0.3253	B7/(b6-b8)	−0.4235
		b7/b8	−0.1748	(b7 + b8)/b6	−0.3050	B6/(b7-b8)	0.2464

Table 3. TOC concentration inversion empirical model parameters.

x1	x2	fTOC(x)	Modle type	Equation	R^2	RMSE
B7	–	TOC	linear	ftoc(x1)=3.517−4.84x1	0.403	0.417
			quadratic	ftoc(x1)=2.965 + 1.946 x 1−17.607x1^2	0.457	0.33
			power	ftoc(x1)=1.396x1^-0.343	0.433	0.37
			exponential	ftoc(x1)=3.684e^-1.964x1	0.454	0.424
B7 × B8	–	TOC	linear	ftoc(x1)=3.086−12.049x1	0.422	0.426
			quadratic	ftoc(x1)=3.130−15.297x1+32.991x1^2	0.424	0.434
			power	ftoc(x1)=1.367 x 1^-0.175	0.427	0.371
			exponential	ftoc(x1)=1.607e^−0.432x1	0.423	0.426
B7	B7 × B8	TOC	multiple regression	ftoc(x1,x2)=2.476x1−17.602x2+2.822	0.456	0.386

Table 4. Comparison of BPNN and quadratic TOC concentration inversion modeling results.

Sampling point	Measured value	Simulation value		Relative error
		BPNN	Quadratic	BPNN	Quadratic
SG3	2.796	2.726	2.96	2.52%	5.88%
SG5	1.664	2.112	1.931	26.91%	16.06%
ZW5	2.5	2.305	2.95	7.79%	17.99%
ZW8	3.12	2.929	2.972	6.12%	4.76%
ZW9	2.338	2.141	2.034	8.41%	13.02%
BQ1	2.656	2.537	2.426	4.48%	8.65%
Average relative error				9.37%	11.06%

Figure 4. Scatterplot of simulated values of BPNN versus quadratic TOC concentration inversion models.

Figure 5. BPNN TOC concentration inversion model simulation results.

Table 5. Correlation of TOC concentration with other water quality parameters.

Water quality parameters	Pearson
Chla	0.271
WT	0.655**
TN	−0.666**
pH	0.136
EC	−0.393*
DO	0.182

Note: *indicates a significant correlation (two-tailed), p < 0.05. **indicates a highly significant correlation (two-tailed), p < 0.01.