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 (R2), 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 (R2 = 0.972, RMSECV = 0.194, RPD = 5.995) and TN (R2 = 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.
This study evaluates the quality of the Advanced Spaceborne Thermal Emission Radiometer-Global Digital Elevation Model version 2 (ASTER GDEM2) in comparison with the previous version (GDEM1) as well as the Shuttle Radar Topographic Mission (SRTM) DEM and topographic-map-derived DEM (Topo-DEM) using inundation area analysis for the projected location of the Karian dam, Indonesia. In addition, the vertical accuracy of each DEM is evaluated using the Real-Time Kinematic differential Global Positioning Systems (RTK-dGPS) data obtained from an intensive geodetic survey. The results of the inundation area analysis show that GDEM2 produced a higher maximum contour level (MCL) (64 m) than did GDEM1 (55 m), and thus, GDME2 has a better quality. In addition, the GDEM2-derived MCL is similar to those produced by SRTM DEM (69 m) and Topo-DEM (62 m). The improvement in the contour level in GDEM2 is believed to be related to the successful removal of voids (artifacts) and anomalies present in GDEM1. However, our RTK-dGPS results show that the vertical accuracy of GDEM2 is much lower than that of GDEM1 and the other DEMs, which is contradictory to the accuracy stated in the GDEM2 validation document. The vertical profiles of all DEMs show that GDEM2 contains a comparatively large number of undulation effects, thereby resulting in higher root mean square error (RMSE) values. These undulation effects may have been introduced during the GDEM2 validation process. Although the results of this study may be site-specific, it is important that they be considered for the improvement of the next GDEM version.
Plant height (PH) is an important agronomical parameter to assess the growth status in upland rice fields. Recently, field-based phenotyping using unmanned aerial vehicles (UAVs) has received increasing attention as a cost-effective, well-suited sensing technology to measure PH. In this study, we evaluated feasibility of a low-cost small UAV for estimating PH in upland rice fields in Laos with a canopy height model (CHM). Images of the upland field, including 501 plots (= 167 accessions × 3 replicates), were captured by a commercial small UAV (DJI Phantom 4) before emergence and in the near-flowering stage to generate digital surface models (DSMs). The CHM was developed from the difference of the DSMs using UAV images obtained before emergence and before flowering. The CHM metrics of each plot were then calculated using 90–99th percentiles and the top 1–10% largest pixel values of CHM and were compared with the manually measured field PH (78.25–189.75 cm). The predictive accuracy was assessed in the 90–99th percentiles and top 1–10% values of CHM metrics with 5-fold cross-validation procedures. Simple linear regression analyses between the field PH and CHM metrics showed that the top 3% CHM metrics had the best correlation with the field PH (R2 = 0.712, root-mean-square error (RMSE) = 9.142 cm, p < 0.001). Cross-validation procedures also confirmed that the top 3% CHM metrics were the best in terms of accuracy for estimating PH, with an error of 6.963% (8.823 cm) error. Graphical abstract
As a proximal soil sensing technique, laboratory visible and near-infrared (Vis-NIR) spectroscopy is a promising tool for the quantitative estimation of soil properties. However, there remain challenges for predicting soil phosphorus (P) content and availability, which requires a reliable model applicable for different land-use systems to upscale. Recently, a one-dimensional convolutional neural network (1D-CNN) corresponding to the spectral information of soil was developed to considerably improve the accuracy of soil property predictions. The present study investigated the predictive ability of a 1D-CNN model to estimate soil available P (oxalate-extractable P; Pox) content in soils by comparing it with partial least squares (PLS) and random forest (RF) regressions using soil samples (n = 318) collected from natural (forest and non-forest) and cultivated (upland and flooded rice fields) systems in Madagascar. Overall, the 1D-CNN model showed the best predictive accuracy (R2 = 0.878) with a highly accurate prediction ability (ratio of performance to the interquartile range = 2.492). Compared to the PLS model, the RF and 1D-CNN models indicated 4.37% and 23.77% relative improvement in root mean squared error values, respectively. Based on a sensitivity analysis, the important wavebands for predicting soil Pox were associated with iron (Fe) oxide, organic matter (OM), and water absorption, which were previously known wavelength regions for estimating P in soil. These results suggest that 1D-CNN corresponding spectral signatures can be expected to significantly improve the predictive ability for estimating soil available P (Pox) from Vis-NIR spectral data. Rapid and accurate estimation of available P content in soils using our results can be expected to contribute to effective fertilizer management in agriculture and the sustainable management of ecosystems. However, the 1D-CNN model will require a large dataset to extend its applicability to other regions of Madagascar. Thus, further updates should be tested in future studies using larger datasets from a wide range of ecosystems in the tropics.