Rei Sonobe

Random forest classification of crop type using multi-temporal TerraSAR-X dual-polarimetric data

The classification maps are required for the management and the estimation of agricultural disaster compensation; however, those techniques have yet to be established. Some supervised learning models may allow accurate classification. In this study, the Random Forest (RF) classifier and the classification and regression tree (CART) were applied to evaluate the potential of multi-temporal TerraSAR-X dual-polarimetric data, on the StripMap mode, for the classification of crop type. Furthermore, comparisons of the two algorithms and polarizations were carried out. In the study area, beans, beet, grasslands, maize, potato and winter wheat were cultivated, and these crop types were classified using the data set acquired in 2009. The classification results of RF were superior to those of CART, and the overall accuracies were 0.91–0.93.

Parameter tuning in the support vector machine and random forest and their performances in cross-and same-year crop classification using TerraSAR-X

This article describes the comparison of three different classification algorithms for mapping crops in Hokkaido, Japan, using TerraSAR-X data. In the study area, beans, beets, grasslands, maize, potatoes, and winter wheat were cultivated. Although classification maps are required for both management and estimation of agricultural disaster compensation, those techniques have yet to be established. Some supervised learning models may allow accurate classification. Therefore, comparisons among the classification and regression tree (CART), the support vector machine (SVM), and random forests (RF) were performed. SVM was the optimum algorithm in this study, achieving an overall accuracy of 89.1% for the same-year classification, which is the classification using the training data in 2009 to classify the test data in 2009, and 78.0% for the cross-year classification, which is the classification using the training data in 2009 to classify the data in 2012.

Assessing the suitability of data from Sentinel-1A and 2A for crop classification

Sentinel-1A C-SAR and Sentinel-2A MultiSpectral Instrument (MSI) provide data applicable to the remote identification of crop type. In this study, six crop types (beans, beetroot, grass, maize, potato, and winter wheat) were identified using five C-SAR images and one MSI image acquired during the 2016 growing season. To assess the potential for accurate crop classification with existing supervised learning models, the four different approaches namely kernel-based extreme learning machine (KELM), multilayer feedforward neural networks, random forests, and support vector machine were compared. Algorithm hyperparameters were tuned using Bayesian optimization. Overall, KELM yielded the highest performance, achieving an overall classification accuracy of 96.8%. Evaluation of the sensitivity of classification models and relative importance of data types using data-based sensitivity analysis showed that the set of VV polarization data acquired on 24 July (Sentinel-1A) and band 4 data (Sentinel-2A) had the greatest potential for use in crop classification.

Assessing the xanthophyll cycle in natural beech leaves with hyperspectral reflectance

The xanthophyll cycle is critical for protecting the photosynthetic apparatus from light-induced oxidative stress. A clear view of the xanthophyll cycle is thus important for understanding abiotic stresses that are closely related to plant growth and reproduction. The epoxidation state (EPS) is well correlated with the photosynthetic radiation use efficiency, and is widely used for assessing the xanthophyll cycle. The hyperspectral index, photochemical reflectance index (PRI), has been claimed to be closely related with the EPS, and offers instantaneous information of photosynthetic activity: its applications are, however, largely limited to herbaceous and coniferous species, and few studies have ever focussed on both sunlit and shaded leaves of deciduous plants. In the present study, we examined the possibility of applying PRI for tracing the xanthophyll cycle in a typical deciduous species (Fagus crenata Blume) as well as other species in a cold-temperate mountainous area. This is based on a series of experiments with only light stress and other inhibited treatments. Furthermore, searching for new hyperspectral indices has also been attempted based on both original and first derivative spectra. Results revealed that PRI had low correlations with the EPS of deciduous leaves, especially for sunlit leaves. As a comparison, the newly identified dD677, 803, a differential type of index using reflectance derivatives at 677 and 803nm, had a much better performance. The robustness of the newly identified index has been confirmed from both inhibitor-treatments and an additional dataset from other deciduous species. The proposed index is hence applicable for tracing the xanthophyll cycle in deciduous species.

Application of the Sahebi model using ALOS/PALSAR and 66.3 cm long surface profile data

Soil moisture is important information for agricultural fields in which erosion of upper soil layers depends upon the soil moisture, and in which the yield depends on soil water content during sowing, growing and harvest periods. Although sensitivity of microwave backscatters to soil moisture is well understood, several factors, such as surface roughness and incidence angle, can interfere with the estimation of soil moisture using Synthetic Aperture Radar (SAR) data. In this letter, we evaluate the influence of these variables using Advanced Land Observing Satellite (ALOS)/Phased Array type L-Band Synthetic Aperture Radar (PALSAR) data and 66.3 cm long surface profile data using the Sahebi model (Sahebi et al. 2003, Estimation of the moisture content of bare soil from RADASAT-1 SAR using simple empirical models. International Journal of Remote Sensing, 24, pp. 2575–2583). The model applied in this study has a root mean square error (RMSE) of only 1.34 dB, which suggests that 66.3 cm long surface profile data are effective for characterization of surface roughness effects on backscattering coefficients.

Towards a Universal Hyperspectral Index to Assess Chlorophyll Content in Deciduous Forests

The reflectance properties of leaves are influenced by diverse biochemical components including chlorophyll, one of the key indicators related to plant photosynthesis and plant stress. Although a number of hyperspectral indices have been proposed for quantifying leaf chlorophyll concentrations, their applications are largely restricted to where they were developed and can hardly provide satisfactory results in other cases. In this study, universally applicable hyperspectral indices calculated from both original and first-order derivative spectra were identified for quantifying leaf chlorophyll concentrations in deciduous forests. Using the main criteria of the ratio of performance to deviation (RPD) and the widely applicable information criterion (WAIC), new hyperspectral indices were proposed for quantifying chlorophyll concentrations in four independent datasets. The results revealed that the normalized derivative difference between the green peak (520-540 nm) and the end of the red edge (720-740 nm) were effective. The normalized difference type of index using reflectance derivatives at 522 and 728 nm, dND (522, 728), was the most effective index for quantifying chlorophyll concentrations, with an R2 of 0.807 and a lowest root mean square error of 8.67 μg/cm2, n = 816. This index was also validated based on a simulated dataset generated from the model of PROprietes SPECTrales Version 5 (PROSPECT 5). Its applicability for assessing chlorophyll content in various deciduous forests was hence demonstrated. We foresee its wide application in the future.

Mapping crop cover using multi-temporal Landsat 8 OLI imagery

ABSTRACT Crop classification maps are useful for estimating amounts of crops harvested, which could help address challenges in food security. Remote-sensing techniques are useful tools for generating crop maps. Optical remote sensing is one of the most attractive options because it offers vegetation indices (VIs) with frequent revisits and has adequate spatial and spectral resolution and some data has been distributed free of charge. However, sufficient consideration has not been given to the potential of VIs calculated from Landsat 8 Operational Land Imager (OLI) data. This article describes the use of Landsat 8 OLI data for the classification of crops in Hokkaido, Japan. In addition to reflectance, VIs calculated from simple formulas that consisted of combinations of two or more reflectance wavebands were evaluated, as well as the six components of the Kauth–Thomas transform. The VIs based on shortwave infrared bands (bands 6 or 7) improved classification accuracy, and using a combination of all derived data from Landsat 8 OLI data resulted in an overall accuracy of 94.5% (allocation disagreement = 4.492 and quantity disagreement = 1.017).

An experimental comparison between KELM and CART for crop classification using Landsat-8 OLI data

Abstract The operational land imager (OLI) is the latest instrument in the Landsat series of satellite imagery, which officially began normal operations on 30 May 2013. The OLI includes two bands that are not on the thematic mapper series of sensors aboard Landsat-5 and 7; a cirrus band and a coastal/aerosol band. This paper compares the classification and regression tree and the kernel-based extreme learning machine (KELM) for mapping crops in Hokkaido, Japan, using OLI data, except the cirrus band and the pan band. The OLI data acquired on 8 July 2013 was used for crop classification of beans, beets, grassland, maize, potatoes and winter wheat. The KELM algorithm performed better in this study and achieved overall accuracies of 90.1%. According to the Jeffries–Matusita (J–M) distances, the short wavelength infrared band provides the greater contribution (the highest value was observed for band 6 in OLI data).

Estimating leaf carotenoid contents of shade-grown tea using hyperspectral indices and PROSPECT–D inversion

ABSTRACT Quantifying carotenoid contents has many applications in agriculture, ecology, and health science. Hyperspectral reflectance has been one of the promising tools for this purpose. However, previous studies were based on measurements under relatively low light–stress conditions. Therefore, assessing its robustness by using measurements under various levels of stress is required. In this study, the measurements of reflectance and carotenoid contents were carried out with four shading treatments including open–0%, 35%, 75%, and 90% shading to generate various chlorophyll/carotenoid ratios. Then the performances of 15 published hyperspectral indices and PROSPECT–D inversion were evaluated based on our data set for estimating leaf carotenoid contents. According to the ratio of performance to deviation, RNIR/R510, R720/R521–1, and PROSPECT–D inversion were applicable for this purpose, although calibration of the absorption coefficients was required for PROSPECT–D. Using them, root mean square percentage errors of 4.53–5.46% were achieved. Given that total chlorophyll/carotenoid ratios could be a good indicator for evaluating environmental stress in plants, PROSPECT–D, which also estimates total chlorophyll and anthocyanin contents, could be a strong tool for controlling the qualities of shade-grown tea.

Crop classification from Sentinel-2 derived vegetation indices using ensemble learning

Abstract. The identification and mapping of crops are important for estimating potential harvest as well as for agricultural field management. Optical remote sensing is one of the most attractive options because it offers vegetation indices and some data have been distributed free of charge. Especially, Sentinel-2A, which is equipped with a multispectral sensor (MSI) with blue, green, red, and near-infrared-1 bands at 10 m; red edge 1 to 3, near-infrared-2, and shortwave infrared 1 and 2 at 20 m; and 3 atmospheric bands (band 1, band 9, and band 10) at 60 m, offer some vegetation indices calculated to assess vegetation status. However, sufficient consideration has not been given to the potential of vegetation indices calculated from MSI data. Thus, 82 published indices were calculated and their importance were evaluated for classifying crop types. The two most common classification algorithms, random forests (RF) and support vector machine (SVM), were applied to conduct cropland classification from MSI data. Additionally, super learning was applied for more improvement, achieving overall accuracies of 90.2% to 92.2%. Of the two algorithms applied (RF and SVM), the accuracy of SVM was superior and 89.3% to 92.0% of overall accuracies were confirmed. Furthermore, stacking contributed to higher overall accuracies (90.2% to 92.2%), and significant differences were confirmed with the results of SVM and RF. Our results showed that vegetation indices had the greatest contributions in identifying specific crop types.