Hyperspectral sensing, measuring reflectance over visible to shortwave infrared wavelengths, has enabled the classification and mapping of vegetation at a range of taxonomic scales, often down to the species level. Classification with hyperspectral measurements, acquired by narrow band spectroradiometers or imaging sensors, has generally required some form of spectral feature selection to reduce the dimensionality of the data to a level suitable for the construction of a classification model. Despite the large number of hyperspectral plant classification studies, an in-depth review of feature selection methods and resultant waveband selections has not yet been performed. Here, we present a review of the last 22 years of hyperspectral vegetation classification literature that evaluates the overall waveband selection frequency, waveband selection frequency variation by taxonomic, structural, or functional group, and the influence of feature selection choice by comparing such methods as stepwise discriminant analysis (SDA), support vector machines (SVM), and random forests (RF). This review determined that all characteristics of hyperspectral plant studies influence the wavebands selected for classification. This includes the taxonomic, structural, and functional groups of the target samples, the methods, and scale at which hyperspectral measurements are recorded, as well as the feature selection method used. Furthermore, these influences do not appear to be consistent. Moreover, the considerable variability in waveband selection caused by the feature selectors effectively masks the analysis of any variability between studies related to plant groupings. Additionally, questions are raised about the suitability of SDA as a feature selection method, with it producing waveband selections at odds with the other feature selectors. Caution is recommended when choosing a feature selector for hyperspectral plant classification: We recommend multiple methods being performed. The resultant sets of selected spectral features can either be evaluated individually by multiple classification models or combined as an ensemble for evaluation by a single classifier. Additionally, we suggest caution when relying upon waveband recommendations from the literature to guide waveband selections or classifications for new plant discrimination applications, as such recommendations appear to be weakly generalizable between studies.

Lao People’s Democratic Republic (PDR) has been experiencing significant forest depletion since the 1980s, but there is little evidence to demonstrate the major causes and underlying drivers for the forest cover changes. In this study, we investigated the relationship between forest cover decrease and increase in the south of Lao PDR between 2006 and 2012 and selected physical and socio-economic factors. We used a map of forest cover changes derived from analysis of Landsat ETM+ imagery in 2006 and 2012, together with socio-economic and physical environmental data from the national authorities. The study area has experienced noticeable forest cover changes: both forest decreases and increases were unevenly distributed throughout the region. Logistic regression models were used to test relationships between forest cover decrease or increase and selected physical and socio-economic factors. Forest clearance was associated strongly with elevation, distance to main roads and shifting cultivation practices. Meanwhile, forest cover increase was more likely to correlate with rubber plantations. Native forest and shifting cultivation lands were vulnerable to being converted into rubber plantations. This research provides much-needed information on which to base forestry policy and decision making to minimize and prevent current deforestation, as well as manage potential risks in the future.

MODIS enhanced vegetation index (EVI) and land surface temperature (LST) are key indicators for monitoring vegetation cover changes in broad ecosystems. However, there has been little evaluation of these indices for detecting changes in a range of land covers in tropical regions. In this study, we investigated the characteristics and seasonal responses of LST and EVI for four different land covers in Lao tropical forests: native forest, rubber plantation, mixed wooded/cleared areas and agriculture. We calculated long-term averages of MODIS LST and EVI 16-day time series and compared their monthly transitions over the seven-year period from 2006 to 2012. We also tested whether these indices can be used to classify these four land covers. The findings demonstrate the complex interrelationship of LST and EVI and their monthly transitions for different land covers: they each showed distinctly different intra-annual LST and EVI variations. Native forests have the highest EVI, and the lowest LST throughout the year. In contrast, agricultural areas with little or no vegetation cover have the highest LST. The transition of LST/EVI for the land covers other than native forests showed marked seasonality. Linear discriminant analysis (LDA) showed that there was high overall accuracy of separation of land covers by these indices (86%). The encouraging results indicate that the combined use of MODIS LST and EVI holds promise for improving monitoring of changes in a Lao tropical forest.

New, accurate and generalizable methods are required to transform the ever-increasing amount of raw hyperspectral data into actionable knowledge for applications such as environmental monitoring and precision agriculture. Here, we apply advances in generative deep learning models to produce realistic synthetic hyperspectral vegetation data, whilst maintaining class relationships. Specifically, a Generative Adversarial Network (GAN) is trained using the Cramér distance on two vegetation hyperspectral datasets, demonstrating the ability to approximate the distribution of the training samples. Evaluation of the synthetic spectra shows that they respect many of the statistical properties of the real spectra, conforming well to the sampled distributions of all real classes. Creation of an augmented dataset consisting of synthetic and original samples was used to train multiple classifiers, with increases in classification accuracy seen under almost all circumstances. Both datasets showed improvements in classification accuracy ranging from a modest 0.16% for the Indian Pines set and a substantial increase of 7.0% for the New Zealand vegetation. Selection of synthetic samples from sparse or outlying regions of the feature space of real spectral classes demonstrated increased discriminatory power over those from more central portions of the distributions.

Although the potential of river discharge to support ocean productivity and marine ecosystems is known, the specifics of this relationship are poorly understood in many regions of the world. Global estimates of river flow indicate that river discharge is decreasing due to the increasing fragmentation, extraction and regulation of rivers. This likely means that the contribution of river flow to coastal productivity and water quality is changing, potentially leading to fewer and smaller magnitude ocean fertilisation events. We developed a simple analysis method, based on Earth observation data, to investigate where coastal ocean chlorophyll‐a is most strongly influenced by river discharge. The per‐pixel spatiotemporal correlation technique (implemented using Python) correlates chlorophyll‐a concentration (a proxy for phytoplankton biomass and indicator of primary productivity) from MODIS ocean colour data with river discharge data. The method was tested globally on 11 different rivers discharging into coastal ocean regions. Our findings suggest some of the world's largest river systems, such as the Amazon River, have zones of elevated coastal chl‐a that extend hundreds to thousands of km from the river mouth. These findings suggest the influence of river discharge may have been underestimated in many coastal regions of the world. The method appears more effective for larger river systems discharging to ocean waters with less complex nutrient dynamics and weaker seasonal productivity patterns, most notably in temperate regions. Increasing our understanding of the specific areas influenced by river discharge, and the degree of influence over space and time, is an important step towards the improved river and coastal management. This method will increase the capacity of researchers to monitor how, when and where coastal waters are affected as river discharge continues to change into the future.