Author Affiliations
1College of Resource and Environment Sciences, Xinjiang University, Urumqi, Xinjiang 830046, China2Key Laboratory of Oasis Ecology, Xinjiang University, Urumqi, Xinjiang 830046, China3Key Laboratory of Smart City and Environment Modelling of Higher Education Institute, Xinjiang University, Urumqi, Xinjiang 830046, Chinashow less
Fig. 1. Statistical characteristics of SMC
Fig. 2. Hyperspectral images based on different pretreatments. (a) Three-dimensional image; (b) R; (c) FDR; (d) SDR; (e) CR; (f) A; (g) FDA; (h) SDA
Fig. 3. Spectral curves based on different pretreatments. (a) R; (b) FDR; (c) SDR; (d) CR; (e) A; (f) FDA; (g) SDA
Fig. 4. Characteristic bands selected by different algorithms. (a)-(c) Characteristic bands of R after RF, GBRT, XGBoost screening; (d)-(f) characteristic bands of FDR after RF, GBRT, XGBoost screening; (g)-(i) characteristic bands of SDR after RF, GBRT, XGBoost screening; (J)-(l) characteristic bands of CR after RF, GBRT, XGBoost screening; (m)-(o) characteristic bands of RF, GBRT, XGBoost screening; (p)-(r) characteristic bands of FDA after RF, GBRT, XGBoost screening; (s)-(u) characteristic band of S
Fig. 5. SMC estimation results based on different preferred methods. (a)-(c) SMC estimation effect of R optimized by RF, GBRT and XGBoost; (d)-(f) SMC estimation effect of FDR optimized by RF, GBRT and XGBoost; (g)-(i) SMC estimation effect of SDR optimized by RF, GBRT and XGBoost; (j)-(l) SMC estimation effect of CR optimized by RF, GBRT and XGBoost; (m)-(o) SMC estimation effect of A optimized by RF, GBRT and XGBoost; (p)-(r) SMC estimation effect of FDA optimized by RF, GBRT and XGBoost; (s)-(u) SMC
Fig. 6. Distribution of characteristic bands
Independent variable | Modeling set | Validation set | |
---|
R2 | | RMSE /% | R2 | RMSE /% | RPIQ |
---|
R-RF | 0.690 | 3.307 | 0.694 | 2.068 | 1.682 | R-GBRT | 0.700 | 3.214 | 0.698 | 2.019 | 1.890 | R-XGBoost | 0.653 | 3.440 | 0.657 | 2.230 | 1.410 | FDR-RF | 0.621 | 3.614 | 0.621 | 2.237 | 1.401 | FDR-GBRT | 0.800 | 2.624 | 0.801 | 1.654 | 3.007 | FDR-XGBoost | 0.771 | 2.802 | 0.772 | 1.764 | 2.943 | SDR-RF | 0.712 | 3.132 | 0.712 | 2.065 | 1.895 | SDR-GBRT | 0.744 | 2.973 | 0.745 | 1.90 | 2.724 | SDR-XGBoost | 0.690 | 3.268 | 0.692 | 2.563 | 1.688 | CR-RF | 0.726 | 3.062 | 0.724 | 1.932 | 2.212 | CR-GBRT | 0.681 | 3.312 | 0.680 | 2.202 | 1.436 | CR-XGBoost | 0.688 | 3.276 | 0.689 | 2.322 | 1.483 | A-RF | 0.694 | 3.239 | 0.692 | 2.076 | 1.724 | A-GBRT | 0.685 | 3.280 | 0.688 | 2.191 | 1.437 | A-XGBoost | 0.690 | 3.257 | 0.691 | 2.053 | 1.588 | FDA-RF | 0.842 | 2.434 | 0.843 | 1.454 | 3.115 | FDA-GBRT | 0.890 | 2.024 | 0.890 | 1.337 | 3.490 | FDA-XGBoost | 0.764 | 2.852 | 0.764 | 1.835 | 2.801 | SDA-RF | 0.599 | 3.727 | 0.598 | 2.317 | 1.362 | SDA-GBRT | 0.738 | 2.998 | 0.740 | 1.881 | 2.315 | SDA-XGBoost | 0.860 | 2.285 | 0.861 | 1.632 | 3.238 |
|
Table 1. GWR model of optimal variable SMC under different preferred methods