Skip to main content

Table 5 Accuracy of transpiration rate prediction models based on multiple thermal infrared indices

From: Monitoring of drought stress and transpiration rate using proximal thermal and hyperspectral imaging in an indoor automated plant phenotyping platform

RMSE

R2

MAPE

Selected variables

Method

Stepwise

0.58

0.57

35

ICWSIin, CWSIdev_out, CWSITa_gz, TRgz, ICWSIgz, CTDgz, Ta_out, VPDout, Ta_in, TBS, Tdry

Stepwise: index + env

0.60

0.53

36

CTDout, ICWSIin, CWSIdev_out, CSIout, CWSITa_in, TRout, VPDout, TBS, 502 nm, 520 nm, 576 nm, 727 nm, 746 nm, 766 nm, 823 nm, 920 nm, 979 nm, 1295 nm, 1407 nm, 1445 nm, 1519 nm, 1557 nm, 1893 nm, 1930 nm, 2451 nm, 2525 nm, 2042 nm

Stepwise: index + env + hyp

LASSO

0.47

0.63

32

CTDout, CSIin, CSIout, ICWSIin, ICWSIout, CWSIdev_in, CWSIdev_out, CWSIdev_gz, CWSITa_gz, TBS-Tp, TRin, TRgz, Ta_out,, PARgz, TBS

LASSO: index + env

0.49

0.69

33

CTDin, CTDout, CSIin, ICWSIin, ICWSIout, CWSITa_gz, Tdry—Tp, Ta_out, VPDin, PARggz, 403 nm, 449 nm, 468 nm, 904 nm, 1207 nm, 1213 nm, 2346 nm, 2377 nm, 2395 nm, 2426 nm, 2439 nm, 2445 nm, 2457 nm, 2476 nm, 2482 nm, 2525 nm

LASSO: index + env + hyp

0.64

0.48

41

Tp, 492 nm, 495 nm, 526 nm, 545 nm, 576 nm, 604 nm, 610 nm, 626 nm, 645 nm, 654 nm, 658 nm, 667 nm, 724 nm, 1002 nm, 1006 nm, 1251 nm, 1276 nm, 1382 nm, 1395 nm, 1526 nm, 1893 nm, 2383 nm, 2426 nm, 2457 nm, 2476 nm, 2513 nm

LASSO: Tp + hyp

0.59

0.48

40

492 nm, 508 nm, 545 nm, 576 nm, 658 nm,1238 nm, 1276 nm, 1395 nm, 1887 nm, 2506 nm, 2513 nm

LASSO: hyp

Random forest

0.54

0.57

33

CTDout, CSIin, CSIout, ICWSIin, CWSIdev_in, CWSIdev_out, CWSIdev_gz, CWSITa_gz, TBS-Tp, TRin, TRgz, Ta_out, VPDgz, PARgz, TBS,

RF: index + env

0.59

0.61

33

CTDout, Tdry-Tp, TBS, Ta_out, Tdry, VPDout, PARgz, 486 nm, 2488 nm, 2494 nm, 2500 nm, 2506 nm, 2513 nm, 2519 nm, 2531 nm

RF: index + env + hyp

Partial least squares regression

0.54

0.62

34

CTDout, TRout, CWSITa_out, TBS-Tp, CWSIdev_out, CSIout, Tdry-Tp, CWSITa_gz, CTDin, CWSIin, ICWSIout, TRin, ICWSIin, CWSITa_in, PARgz, TBS, Ta_out, Tdry, Ta_in, Ta_gz, VPDout, VPDin, VPDgz, RHout, RHin, Tp, 483 nm, 486 nm, 489 nm, 492 nm, 495 nm, 502 nm, 505 nm, 508 nm, 511 nm, 514 nm, 517 nm, 2488 nm, 2494 nm, 2506 nm, 2500 nm, 2513 nm, 2519 nm, 2525 nm, 2531 nm,

PLSR: index + env + hyp

  1. Median test RMSE, R2 and MAPE were calculated from 100 bootstrap samples. The data used to create these models consisted of TIR indices (index, Table 1), environmental data (env), hyperspectral indices and the reflectance of the hyperspectral wavelength bands (hyp)