Genetic algorithm‐based partial least squares regression for estimating legume content in a grass‐legume mixture using field hyperspectral measurements

Kawamura, Kensuke; Watanabe, Naoharu; Sakanoue, Seiichi; Lim, Jihyun; Yoshitoshi, Rena

doi:10.1111/grs.12026

Cited by 11 publications

(12 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These results agree with previous results indicating that the most useful information in the Vis-NIR region (400-2400 nm) was less than 20% for predicting forage [18,19] and water parameters [20]. These findings also support previous results showing that the performance of PLS models can be improved through waveband selection.…”

Section: Soil Parametersupporting

confidence: 92%

“…Surface soil samples were collected from a 0-10 cm depth as composites of three to four cores in each field. Within three fields, sub-surface samples (10-20 cm depth in a field; [10][11][12][13][14][15][16][17][18][19][20][20][21][22][23][24][25][26][27][28][29][30], and 30-40 cm depth in two fields) were also collected. Thus, 62 soil samples were obtained.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Vis-NIR Spectroscopy and PLS Regression with Waveband Selection for Estimating the Total C and N of Paddy Soils in Madagascar

et al. 2017

Self Cite

View full text Add to dashboard Cite

Visible and near-infrared (Vis-NIR) diffuse reflectance spectroscopy with partial least squares (PLS) regression is a quick, cost-effective, and promising technology for predicting soil properties. The advantage of PLS regression is that all available wavebands can be incorporated in the model, while earlier studies indicate that PLS models include redundant wavelengths, and selecting specific wavebands can refine PLS analyses. This study evaluated the performance of PLS regression with waveband selection using Vis-NIR reflectance spectra to estimate the total carbon (TC) and total nitrogen (TN) in soils collected mainly from the surface of upland and lowland rice fields in Madagascar (n = 59; after outliers were removed). We used iterative stepwise elimination-based PLS (ISE-PLS) to estimate soil TC and TN and compared the predictive ability with standard full-spectrum PLS (FS-PLS). The predictive abilities were assessed using the coefficient of determination (R 2 ), the root mean squared error of cross-validation (RMSECV), and the residual predictive deviation (RPD). Overall, ISE-PLS using first derivative reflectance (FDR) showed a better predictive accuracy than ISE-PLS for both TC (R 2 = 0.972, RMSECV = 0.194, RPD = 5.995) and TN (R 2 = 0.949, RMSECV = 0.019, RPD = 4.416) in the soil of Madagascar. The important wavebands for estimating TC (12.59% of all wavebands) and TN (3.55% of all wavebands) were selected from all 2001 wavebands over the 400-2400 nm range using ISE-PLS. These findings suggest that ISE-PLS based on Vis-NIR diffuse reflectance spectra can be used to estimate soil TC and TN contents in Madagascar with an improved predictive accuracy.

show abstract

Section: Soil Parametersupporting

confidence: 92%

Section: Introductionmentioning

confidence: 99%

Vis-NIR Spectroscopy and PLS Regression with Waveband Selection for Estimating the Total C and N of Paddy Soils in Madagascar

et al. 2017

Self Cite

View full text Add to dashboard Cite

show abstract

“…Considering the saturation problem, the forage parameters were compared between the normal and logarithmic forms (ln BM, ln LAI and CP mass ) (Watanabe et al 2014). The performance of the models was evaluated by comparing the differences in the cross-validated (leave-one-out cross-validation method) coefficient of determination (R 2 cv ) (Kawamura et al 2013) and residual prediction (RPD) values (Williams 2001). The RPD value represents the standard deviation (SD) of the forage parameter data divided by the standard error of cross-validation (SECV) and provides a comparison of the performance of all of the calibrations irrespective of the units of the investigated parameters (Park et al 1997).…”

Section: Discussionmentioning

confidence: 99%

Evaluating a hand‐held crop‐measuring device for estimating the herbage biomass, leaf area index and crude protein content in an Italian ryegrass field

et al. 2015

Self Cite

View full text Add to dashboard Cite

The objective of this study is to evaluate the ability of a newly developed hand‐held crop‐measuring device and vegetation indices (VIs) to estimate the herbage biomass (BM), leaf area index (LAI) and forage crude protein mass (CPmass) in an Italian ryegrass (Lolium multiflorum Lam.) field, Japan. The device uses bi‐directional passive sensors (550, 650 and 880 nm) upward and downward to overcome the major drawback of optical remote sensing as influenced by weather conditions. The canopy reflectance and plant sample data were collected 11 times during two winter growing seasons in 2010–11 and 2011–12. Seven VIs were compared to estimate the forage parameters in the normal and logarithmic forms. The predictive ability of the VIs was assessed by the cross‐validated coefficient of determination (R2cv) and the residual prediction (RPD) values. Acceptable RPD values (>1.5) were found in most of the log‐transformed forage parameters with all of the VIs but not in most of the normal‐form. The highest R2cv and RPD were obtained in the normalized difference vegetation index (NDVI) for the ln BM (R2cv = 0.76, RPD = 2.04) and ln LAI (R2cv = 0.80, RPD = 2.25), and the highest modified soil‐adjusted vegetation index (MSAVI) was obtained for the ln CPmass (R2cv = 0.78, RPD = 2.16). By evaluating the limitation of NDVI or MSAVI sensitivity with ranges of cumulative data, the log‐transformed forage parameters showed good R2cv values in the ln BM (0.76–0.79), ln LAI (0.80–0.84) and ln CPmass (0.71–0.84) with acceptable RPD values throughout most of the range, whereas in the normal‐form, the NDVI or MSAVI was saturated at moderate‐high BM, LAI or CPmass, causing the R2cv and RPD values to decrease with increasing plant parameters. These results suggest that this device is applicable in log‐transformed BM, LAI and CPmass throughout the growing season without cloud effect.

show abstract

“…Also, the wavebands selected for the reflectance equations do not describe the optical properties of the forages but instead describe how the forages were distinguished from others. Biewer et al (2009) and Kawamura et al (2013) concluded that using spectroscopy to predict legume content in grass-legume mixtures became more difficult as the grass dominance increased, thereby confounding potential spectral differences with biomass effects. Future work should include new software that characterizes leaf shape (straight or curved edges) as seen in McRoberts et al (2016).…”

Section: Model Evaluationmentioning

confidence: 99%

Nondestructive Determination of Legume Content in Grass‐Legume Pastures

Baxter¹,

West²,

Brown³

et al. 2017

Crop Forage & Turfgrass Mgmt

View full text Add to dashboard Cite

Core Ideas Procedures were more successful when forage species were distinct hues of green. Overall, predictive accuracy of PowerPoint model was best in Old World bluestem–legume pasture. Only the visual model showed potential in tall wheatgrass–alfalfa pasture. To better manage forage productivity and quality, it is important to accurately and objectively monitor legume content in grass‐legume pastures. Current nondestructive methods are too subjective or labor intensive. Recent research has led to developments in digital image analysis (DIA) and remote sensing techniques useful in pasture research and management. Four nondestructive sampling techniques were compared with traditional botanical hand separations to determine their ability to predict legume content in pastures of tall wheatgrass [Thinopyrum ponticum (Podp.); TW] and alfalfa (Medicago sativa L.): TW‐alfalfa; and in pastures of Old World bluestem [Bothriochloa bladhii (Retz) S.T. Blake; OWB], alfalfa, and yellow sweetclover (Melilotus officinalis L.; YSC): OWB‐legume. Each sampling procedure better estimated and predicted legume content in OWB‐legume pasture (R2cal = 0.40 to 0.78; R2pred = 0.40 to 0.78) than in TW‐alfalfa (R2cal = 0.02 to 0.61; R2pred = 0.08 to 0.61). Overall, the best predictive accuracy was obtained with the PowerPoint model in OWB‐legume pasture. Only the visual model showed potential in TW‐alfalfa because these two species were difficult to distinguish on‐screen by DIA methods or using spectral thresholds. The use of nondestructive sampling techniques to measure legume content is promising for mixed pastures when the grass and legume components are distinct hues of green. Although the need for more research is recognized, the ultimate goal is to apply these findings to automated scanners that offer producers rapid measurements of legume content in mixed pastures.

show abstract

Genetic algorithm‐based partial least squares regression for estimating legume content in a grass‐legume mixture using field hyperspectral measurements

Cited by 11 publications

References 29 publications

Vis-NIR Spectroscopy and PLS Regression with Waveband Selection for Estimating the Total C and N of Paddy Soils in Madagascar

Vis-NIR Spectroscopy and PLS Regression with Waveband Selection for Estimating the Total C and N of Paddy Soils in Madagascar

Evaluating a hand‐held crop‐measuring device for estimating the herbage biomass, leaf area index and crude protein content in an Italian ryegrass field

Nondestructive Determination of Legume Content in Grass‐Legume Pastures

Contact Info

Product

Resources

About