Description: Farmland soil environmental quality is important for farmland management. To precisely classify the environmental quality grades of farmland soil, additional samples may be required for multistage sampling or supplementary investigations. Compared with the sampling optimization methods used for mapping or estimating global means, environmental quality grade classifications are primarily focused on estimating the relationships between the values of unsampled locations and the thresholds that classify the environment quality grades. Such classifications must use a sampling layout optimization method to distribute additional sampling units into areas with a high risk of misclassification. To resolve such problems, this paper provides an additional sampling layout optimization method that initially develops a classification error index by building a multi-Gaussian model with the predicted values and error variances of unsampled locations and then calculates the probability of a threshold value occurring in the standardized Gaussian distribution. The average error indexes of all locations in the study area are then set as the objectivity function of the additional sampling layout optimization, and the spatial simulated annealing is adopted to obtain the optimized sampling layout by minimizing the objectivity function. The performance of the error index sampling layout optimization method was demonstrated in a case study using chromium concentration data for Hunan Province, China. The results showed that the additional samples generated by the proposed method produce lower and more stable classification error rates than the minimization of the mean of the shortest distances and spatially random sample methods. The  proposed method can be used to improve the efficiency of additional sampling for  environmental quality grade classifications of farmland soil.

Description: Efficient water management in agriculture requires an accurate estimation of evapotranspiration (ET). Even though local measurements can be used to estimate the components of the surface energy balance, these values cannot be extrapolated to large areas due to the heterogeneity and complexity of agricultural and natural land surfaces and the dynamic nature of their heat processes. This extrapolation can be done by using satellite imagery, which provides information in the infrared thermal band; however, this band is not available in most current operational remote sensors. Our work hypothesis is that it is possible to generate a spatially distributed estimation of $text{ET}_{d}$ without thermal band by using nonparametric models as regression random forest models (RRFM). Six Landsat-7 scenes were used to generate the RRFM. Results were evaluated by comparing the values of $text{ET}_{d}$ provided by RRFM with that obtained using the surface energy balance model. It has been shown that the results generated by RRFM present a good agreement with METRIC (mapping ET at high resolution using internalized calibration) results, both quantitatively and qualitatively, especially for agricultural vegetation and forest land covers. Moreover, it has been detected that the RRFM estimation quality depends on the meteorological conditions on the days previous to the satellite register. It can be concluded that the $text{ET}_{d}$ estimated by the RRFM would be feasible for real applications when the thermal band is not available.

Description: Information on crop types derived from remotely sensed images provides valuable input for many applications such as crop growth modeling and yield forecasting. In this paper, a random forest (RF) classifier was used for crop classification using multispectral RapidEye imagery over two study sites, one in north-eastern China and one in eastern Ontario, Canada. Both vegetation indices (VIs) and textural features were derived from the RapidEye imagery and used for classification. A total of 20 VIs, categorized into two groups with and without the red edge (RE) band in an index, were calculated. A total of eight types of textural features were derived using four different window sizes from both the RE and the near-infrared bands. To reduce redundancies among the VIs and textural features, feature selection using the principal component analysis, correlation analysis, and stepwise discriminant analysis was performed. Results showed that the overall classification accuracy was improved by ∼7% when the RE indices were combined with the five spectral bands in classification, as compared with that using the five bands alone. When textural information was included, the overall classification accuracy increased by ∼6% compared with that using the band reflectance alone. Furthermore, when all the features (band reflectance, VIs, and texture) were used, the overall classification accuracy increased by ∼12% compared with that using only the band reflectance. The RF importance measures showed that the RE reflectance was important for classification, as indicated by the high importance for the triangular vegetation index, transformed chlorophyll absorption in reflectance index, and green-rededge normalized difference vegetation index. The gray-level co-occurrence matrix mean is the most useful for classification among the textural features. The study provides a means t- feature extraction and selection for crop classification from remote sensing imagery.

Description: Foliar chemical constituents are important indicators for understanding vegetation growing status and ecosystem functionality. Provided the noncontact and nondestructive traits, the hyperspectral analysis is a superior and efficient method for deriving these parameters. In practice, the spectral noise issue significantly impacts the performance of the hyperspectral retrieving system. To systematically investigate this issue, by introducing varying levels of noise to spectral signals, an assessment on noise-resistant capability of spectral features and models for retrieving concentrations of chlorophyll, carotenoids, and leaf water content was conducted. Given the continuous wavelet analysis (CWA) showed superior performance in extracting critical information associating plants biophysical and biochemical status in recent years, both wavelet features (WFs) and some conventional features (CFs) were chosen for the test. Two datasets including a leaf optical properties experiment dataset $(n,= ,330)$ , and a corn leaf spectral experiment dataset $(n,= ,213)$ were used for analysis and modeling. The results suggested that the WFs had stronger correlations with all leaf chemical parameters than the CFs. According to an evaluation by decay rate of retrieving error that indicates noise-resistant capability, both WFs and CFs exhibited strong resistance to spectral noise. Particularly for WFs, the noise-resistant capability is relevant to the scale of the features. Based on the identified spectral features, both univariate and multivariate retrieving models were established and achieved satisfactory accuracies. Synthesizing the retrieving accuracy, noise resistivity, and model's complexity, the optimal univariate WF-models were recommended in practice for retrieving leaf chemical parameters.

Description: Nematodes are a serious issue for coffee cultivation in Brazil. Because root infection by nematodes induces spectral variation in leaves and defines a unique spatial configuration in the cultivation field, the aim of this study was to use biophysical parameters and remote sensing data to discriminate and map healthy, moderately infected, and severely infected coffee plants. An experimental area in southern Minas Gerais State, in which the occurrence of nematodes was certified, was selected, and biophysical and spectral measurements of the leaves were made. Hyperspectral data were also used in a band simulation of the RapidEye sensor to identify the most sensitive spectral ranges for pathogen discrimination in coffee plants. These bands, plus a normalized difference vegetation index image, were used for a multispectral classification of healthy and nematode-infected areas. None of the biophysical parameters efficiently discriminated the leaves of healthy and infected plants, but the band simulation indicated that red, red edge, and near infrared spectral ranges were complementary to the discrimination of healthy coffee plants and the two levels of infection. The multispectral classification defined the spatial distribution of healthy, moderately infected, and severely infected coffee plants, with an overall accuracy of 78% and Kappa coefficient of 0.71. Considering the degree of uncertainty and high cost involved in conventional detection of soil parasites, the levels of accuracy achieved were adequate.

Description: Topology adaptive boundary extraction for water in satellite images using parametric snakes remains challenging in the domain of image segmentation due to the inability of topology inflexible parametric snakes (TIPS) to split and merge curves and the limited accuracy or inefficiency of topology flexible parametric snakes (TFPS). Based on the balloon snake (B-snake) method, an outstanding representative of TIPS, this paper proposed the automatic B-snake (AB-snake) method capable of automatic initialization and topology transformation to overcome the drawbacks of nonautomatic initialization and topology inflexibility of the B-snake method while ensuring both high accuracy and efficiency of TFPS. Six GF-1 wide field view satellite images covering different inner island numbers, water body sizes, boundary complexities, boundary clarities, and background complexities were used as experimental images. To test its feasibility and advantages, the extracted results of the AB-snake method were compared with those of the B-snake and the orthogonal topology adaptive snake (OT-snake) methods, a newer representative of TFPS. The results demonstrated that the AB-snake method could effectively implement automatic initialization and topology transformation, successfully meeting its design objectives. Four indices, including correctness, completeness, area overlap measure, and efficiency, were adopted to quantitatively assess the performances of the B-snake, the AB-snake, and the OT-snake methods. The results suggested that the proposed AB-snake method not only significantly outperformed the B-snake method in topology collision handling but also demonstrated great advantages over the OT-snake method in accuracy. The applicability, sensitivity, parameter selection, and limitations of the AB-snake method were also elaborated.