Table 1 Performance comparison of models for TLI retrieval based on S2, GF-1, HJ-2, and PS datasets

From: Resolving inherent constraints in eutrophication monitoring of small lakes using multi-source satellites and machine learning

Sensor

Model

Training set

Test set

Time (s)

NRMSE

Bias

UMAP

R

NRMSE

Bias

UMAP

R

S2

CBR

0.12

0.01

0.05

0.90

0.14

0.15

0.05

0.80

1.40

XGBR

0.06

0.00

0.02

0.97

0.15

0.10

0.06

0.74

0.12

LR

0.19

0.00

0.08

0.64

0.19

0.33

0.08

0.63

0.02

TPFNR

0.12

0.09

0.05

0.89

0.16

0.26

0.06

0.81

68.51

GF-1

CBR

0.09

−0.06

0.03

0.96

0.13

−0.80

0.04

0.94

0.97

XGBR

0.03

0.00

0.01

1.00

0.12

−0.11

0.04

0.91

0.09

LR

0.17

0.00

0.05

0.83

0.21

−0.77

0.06

0.81

0.02

TPFNR

0.13

−0.04

0.03

0.91

0.13

0.15

0.04

0.89

30.12

HJ-2

CBR

0.09

0.01

0.03

0.95

0.16

0.18

0.05

0.83

1.08

XGBR

0.02

0.00

0.01

1.00

0.16

0.66

0.04

0.84

0.12

LR

0.19

0.00

0.07

0.67

0.23

-0.58

0.08

0.63

0.01

TPFNR

0.13

0.05

0.05

0.88

0.17

0.08

0.06

0.80

20.71

PS

CBR

0.12

0.04

0.03

0.91

0.19

−0.09

0.04

0.64

0.84

XGBR

0.08

0.00

0.02

0.97

0.21

−0.10

0.05

0.62

0.06

LR

0.18

0.00

0.04

0.48

0.24

-0.09

0.04

0.47

0.01

TPFNR

0.20

−0.01

0.05

0.62

0.24

0.11

0.03

0.50

17.36