Habitat suitability modelling of an invasive plant with advanced remote sensing data

Earth Observation for Monitoring, Reporting, and Verification within Environmental Land Management Policy

The main aim of the new agricultural scheme, Environmental Land Management, in England is to reward landowners based on their provision of ‘public goods’ while achieving the goals of the 25 Year Environment Plan and commitment to net zero emission by 2050. Earth Observation (EO) satellites appear to offer an unprecedented opportunity in the process of monitoring, reporting, and verification (MRV) of this scheme. In this study, we worked with ecologists to determine the habitat–species relationships for five wildlife species in the Surrey Hills ‘Area of Outstanding Natural Beauty’ (AONB), and this information was used to examine the extent to which EO satellite imagery, particularly very high resolution (VHR) imagery, could be used for habitat assessment, via visual interpretation and automated methods. We show that EO satellite products at 10 m resolution and other geospatial datasets enabled the identification and location of broadly suitable habitat for these species and the use of VHR imagery (at 1–4 m spatial resolution) allowed valuable insights for remote assessment of habitat qualities and quantity. Hence, at a fine scale, we obtained additional habitats such as scrub, hedges, field margins, woodland and tree characteristics, and agricultural practices that offer an effective source of information for sustainable land management. The opportunities and limitations of this study are discussed, and we conclude that there is considerable scope for it to offer valuable information for land management decision-making and as support and evidence for MRV for incentive schemes.

Spatially explicit removal strategies increase the efficiency of invasive plant species control

Effective management strategies are needed to control expansion of invasive alien plant species and attenuate economic and ecological impacts. While previous theoretical studies have assessed optimal control strategies that balance economic costs and ecological benefits, less attention has been paid to the ways in which the spatial characteristics of individual patches may mediate the effectiveness of management strategies. We developed a spatially explicit cellular automaton model for invasive species spread, and compared the effectiveness of seven control strategies. These control strategies used different criteria to prioritize the removal of invasive species patches from the landscape. The different criteria were related to patch size, patch geometry, and patch position within the landscape. Effectiveness of strategies was assessed for both seed dispersing and clonally expanding plant species. We found that, for seed-dispersing species, removal of small patches and removal of patches that are isolated within the landscape comprised relatively effective control strategies. For clonally expanding species, removal of patches based on their degree of isolation and their geometrical properties comprised relatively effective control strategies. Subsequently, we parameterized the model to mimic the observed spatial distribution of the invasive species Antigonon leptopus on St. Eustatius (northern Caribbean). This species expands clonally and also disperses via seeds, and model simulations showed that removal strategies focusing on smaller patches that are more isolated in the landscape would be most effective and could increase the effectiveness of a 10-yr control strategy by 30-90%, as compared to random removal of patches. Our study emphasizes the potential for invasive plant species management to utilize recent advances in remote sensing, which enable mapping of invasive species at the high spatial resolution needed to quantify patch geometries. The presented results highlight how this spatial information can be used in the design of more effective invasive species control strategies.

Tree Species Classification of Forest Stands Using Multisource Remote Sensing Data

The spatial distribution of forest stands is one of the fundamental properties of forests. Timely and accurately obtained stand distribution can help people better understand, manage, and utilize forests. The development of remote sensing technology has made it possible to map the distribution of tree species in a timely and accurate manner. At present, a large amount of remote sensing data have been accumulated, including high-spatial-resolution images, time-series images, light detection and ranging (LiDAR) data, etc. However, these data have not been fully utilized. To accurately identify the tree species of forest stands, various and complementary data need to be synthesized for classification. A curve matching based method called the fusion of spectral image and point data (FSP) algorithm was developed to fuse high-spatial-resolution images, time-series images, and LiDAR data for forest stand classification. In this method, the multispectral Sentinel-2 image and high-spatial-resolution aerial images were first fused. Then, the fused images were segmented to derive forest stands, which are the basic unit for classification. To extract features from forest stands, the gray histogram of each band was extracted from the aerial images. The average reflectance in each stand was calculated and stacked for the time-series images. The profile curve of forest structure was generated from the LiDAR data. Finally, the features of forest stands were compared with training samples using curve matching methods to derive the tree species. The developed method was tested in a forest farm to classify 11 tree species. The average accuracy of the FSP method for ten performances was between 0.900 and 0.913, and the maximum accuracy was 0.945. The experiments demonstrate that the FSP method is more accurate and stable than traditional machine learning classification methods.

Phenology-Based Mapping of an Alien Invasive Species Using Time Series of Multispectral Satellite Data: A Case-Study with Glossy Buckthorn in Québec, Canada

Glossy buckthorn (Frangula alnus Mill.) is an alien species in Canada that is invading many forested areas. Glossy buckthorn has impacts on the biodiversity and productivity of invaded forests. Currently, we do not know much about the species’ ecology and no thorough study of its distribution in temperate forests has been performed yet. As is often the case with invasive plant species, the phenology of glossy buckthorn differs from that of other indigenous plant species found in invaded communities. In the forests of eastern Canada, the main phenological difference is a delay in the shedding of glossy buckthorn leaves, which occurs later in the fall than for other indigenous tree species found in that region. Therefore, our objective was to use that phenological characteristic to map the spatial distribution of glossy buckthorn over a portion of southern Québec, Canada, using remote sensing-based approaches. We achieved this by applying a linear temporal unmixing model to a time series of the normalized difference vegetation index (NDVI) derived from Landsat 8 Operational Land Imager (OLI) images to create a map of the probability of the occurrence of glossy buckthorn for the study area. The map resulting from the temporal unmixing model shows an agreement of 69% with field estimates of glossy buckthorn occurrence measured in 121 plots distributed over the study area. Glossy buckthorn mapping accuracy was limited by evergreen species and by the spectral and spatial resolution of the Landsat 8 OLI.

Multispectral Approach for Identifying Invasive Plant Species Based on Flowering Phenology Characteristics

Invasive plant species (IPS) are the second biggest threat to biodiversity after habitat loss. Since the spatial extent of IPS is essential for managing the invaded ecosystem, the current study aims at identifying and mapping the aggressive IPS of Acacia salicina and Acacia saligna, to understand better the key factors influencing their distribution in the coastal plain of Israel. This goal was achieved by integrating airborne-derived hyperspectral imaging and multispectral earth observation for creating species distribution maps. Hyperspectral data, in conjunction with high spatial resolution species distribution maps, were used to train the multispectral images at the species level. We incorporated a series of statistical models to classify the IPS location and to recognize their distribution and density. We took advantage of the phenological flowering stages of Acacia trees, as obtained by the multispectral images, for the support vector machine classification procedure. The classification yielded an overall Kappa coefficient accuracy of 0.89. We studied the effect of various environmental and human factors on IPS density by using a random forest machine learning model, to understand the mechanisms underlying successful invasions, and to assess where IPS have a higher likelihood of occurring. This algorithm revealed that the high density of Acacia most closely related to elevation, temperature pattern, and distances from rivers, settlements, and roads. Our results demonstrate how the integration of remote-sensing data with different data sources can assist in determining IPS proliferation and provide detailed geographic information for conservation and management efforts to prevent their future spread.

Machine Learning Techniques for Tree Species Classification Using Co-Registered LiDAR and Hyperspectral Data

The use of light detection and ranging (LiDAR) techniques for recording and analyzing tree and forest structural variables shows strong promise for improving established hyperspectral-based tree species classifications; however, previous multi-sensoral projects were often limited by error resulting from seasonal or flight path differences. The National Aeronautics and Space Administration (NASA) Goddard’s LiDAR, hyperspectral, and thermal imager (G-LiHT) is now providing co-registered data on experimental forests in the United States, which are associated with established ground truths from existing forest plots. Free, user-friendly machine learning applications like the Orange Data Mining Extension for Python recently simplified the process of combining datasets, handling variable redundancy and noise, and reducing dimensionality in remotely sensed datasets. Neural networks, CN2 rules, and support vector machine methods are used here to achieve a final classification accuracy of 67% for dominant tree species in experimental plots of Howland Experimental Forest, a mixed coniferous–deciduous forest with ten dominant tree species, and 59% for plots in Penobscot Experimental Forest, a mixed coniferous–deciduous forest with 15 dominant tree species. These accuracies are higher than those produced using LiDAR or hyperspectral datasets separately, suggesting that combined spectral and structural data have a greater richness of complementary information than either dataset alone. Using greatly simplified datasets created by our dimensionality reduction methodology, machine learner performance remains comparable or higher to that using the full dataset. Across forests, the identification of shared structural and spectral variables suggests that this methodology can successfully identify parameters with high explanatory power for differentiating among tree species, and opens the possibility of addressing large-scale forestry questions using optimized remote sensing workflows.

Strength in Numbers: Combining Multi-Source Remotely Sensed Data to Model Plant Invasions in Coastal Dune Ecosystems

A common feature of most theories of invasion ecology is that the extent and intensity of invasions is driven by a combination of drivers, which can be grouped into three main factors: propagule pressure (P), abiotic drivers (A) and biotic interactions (B). However, teasing apart the relative contribution of P, A and B on Invasive Alien Species (IAS) distributions is typically hampered by a lack of data. We focused on Mediterranean coastal dunes as a model system to test the ability of a combination of multi-source Remote Sensing (RS) data to characterize the distribution of five IAS. Using generalized linear models, we explored and ranked correlates of P, A and B derived from high-resolution optical imagery and three-dimensional (3D) topographic models obtained from LiDAR, along two coastal systems in Central Italy (Lazio and Molise Regions). Predictors from all three factors contributed significantly to explaining the presence of IAS, but their relative importance varied among the two Regions, supporting previous studies suggesting that invasion is a context-dependent process. The use of RS data allowed us to characterize the distribution of IAS across broad, regional scales and to identify coastal sectors that are most likely to be invaded in the future.

Toward an integrated approach to crop production and pollination ecology through the application of remote sensing

Insect pollinators provide an essential ecosystem service by transferring pollen to crops and native vegetation. The extent to which pollinator communities vary both spatially and temporally has important implications for ecology, conservation and agricultural production. However, understanding the complex interactions that determine pollination service provisioning and production measures over space and time has remained a major challenge. Remote sensing technologies (RST), including satellite, airborne and ground based sensors, are effective tools for measuring the spatial and temporal variability of vegetation health, diversity and productivity within natural and modified systems. Yet while there are synergies between remote sensing science, pollination ecology and agricultural production, research communities have only recently begun to actively connect these research areas. Here, we review the utility of RST in advancing crop pollination research and highlight knowledge gaps and future research priorities. We found that RST are currently used across many different research fields to assess changes in plant health and production (agricultural production) and to monitor and evaluate changes in biodiversity across multiple landscape types (ecology and conservation). In crop pollination research, the use of RST are limited and largely restricted to quantifying remnant habitat use by pollinators by ascertaining the proportion of, and/or isolation from, a given land use type or local variable. Synchronization between research fields is essential to better understand the spatial and temporal variability in pollinator dependent crop production. RST enable these applications to be scaled across much larger areas than is possible with field-based methods and will facilitate large scale ecological changes to be detected and monitored. We advocate greater use of RST to better understand interactions between pollination, plant health and yield spatial variation in pollinator dependent crops. This more holistic approach is necessary for decision-makers to improve strategies toward managing multiple land use types and ecosystem services.

Contemporary Remotely Sensed Data Products Refine Invasive Plants Risk Mapping in Data Poor Regions

Invasive weeds are a serious problem worldwide, threatening biodiversity and damaging economies. Modeling potential distributions of invasive weeds can prioritize locations for monitoring and control efforts, increasing management efficiency. Forecasts of invasion risk at regional to continental scales are enabled by readily available downscaled climate surfaces together with an increasing number of digitized and georeferenced species occurrence records and species distribution modeling techniques. However, predictions at a finer scale and in landscapes with less topographic variation may require predictors that capture biotic processes and local abiotic conditions. Contemporary remote sensing (RS) data can enhance predictions by providing a range of spatial environmental data products at fine scale beyond climatic variables only. In this study, we used the Global Biodiversity Information Facility (GBIF) and empirical maximum entropy (MaxEnt) models to model the potential distributions of 14 invasive plant species across Southeast Asia (SEA), selected from regional and Vietnam's lists of priority weeds. Spatial environmental variables used to map invasion risk included bioclimatic layers and recent representations of global land cover, vegetation productivity (GPP), and soil properties developed from Earth observation data. Results showed that combining climate and RS data reduced predicted areas of suitable habitat compared with models using climate or RS data only, with no loss in model accuracy. However, contributions of RS variables were relatively limited, in part due to uncertainties in the land cover data. We strongly encourage greater adoption of quantitative remotely sensed estimates of ecosystem structure and function for habitat suitability modeling. Through comprehensive maps of overall predicted area and diversity of invasive species, we found that among lifeforms (herb, shrub, and vine), shrub species have higher potential invasion risk in SEA. Native invasive species, which are often overlooked in weed risk assessment, may be as serious a problem as non-native invasive species. Awareness of invasive weeds and their environmental impacts is still nascent in SEA and information is scarce. Freely available global spatial datasets, not least those provided by Earth observation programs, and the results of studies such as this one provide critical information that enables strategic management of environmental threats such as invasive species.

Invasive Shrub Mapping in an Urban Environment from Hyperspectral and LiDAR-Derived Attributes

Proactive management of invasive species in urban areas is critical to restricting their overall distribution. The objective of this work is to determine whether advanced remote sensing technologies can help to detect invasions effectively and efficiently in complex urban ecosystems such as parks. In Surrey, BC, Canada, Himalayan blackberry (Rubus armeniacus) and English ivy (Hedera helix) are two invasive shrub species that can negatively affect native ecosystems in cities and managed urban parks. Random forest (RF) models were created to detect these two species using a combination of hyperspectral imagery, and light detection and ranging (LiDAR) data. LiDAR-derived predictor variables included irradiance models, canopy structural characteristics, and orographic variables. RF detection accuracy ranged from 77.8 to 87.8% for Himalayan blackberry and 81.9 to 82.1% for English ivy, with open areas classified more accurately than areas under canopy cover. English ivy was predicted to occur across a greater area than Himalayan blackberry both within parks and across the entire city. Both Himalayan blackberry and English ivy were mostly located in clusters according to a Local Moran's I analysis. The occurrence of both species decreased as the distance from roads increased. This study shows the feasibility of producing highly accurate detection maps of plant invasions in urban environments using a fusion of remotely sensed data, as well as the ability to use these products to guide management decisions.

Pixel color clustering of multi-temporally acquired digital photographs of a rice canopy by luminosity-normalization and pseudo-red-green-blue color imaging

Red-green-blue (RGB) channels of RGB digital photographs were loaded with luminosity-adjusted R, G, and completely white grayscale images, respectively (RGwhtB method), or R, G, and R + G (RGB yellow) grayscale images, respectively (RGrgbyB method), to adjust the brightness of the entire area of multi-temporally acquired color digital photographs of a rice canopy. From the RGwhtB or RGrgbyB pseudocolor image, cyan, magenta, CMYK yellow, black, L*, a*, and b* grayscale images were prepared. Using these grayscale images and R, G, and RGB yellow grayscale images, the luminosity-adjusted pixels of the canopy photographs were statistically clustered. With the RGrgbyB and the RGwhtB methods, seven and five major color clusters were given, respectively. The RGrgbyB method showed clear differences among three rice growth stages, and the vegetative stage was further divided into two substages. The RGwhtB method could not clearly discriminate between the second vegetative and midseason stages. The relative advantages of the RGrgbyB method were attributed to the R, G, B, magenta, yellow, L*, and a* grayscale images that contained richer information to show the colorimetrical differences among objects than those of the RGwhtB method. The comparison of rice canopy colors at different time points was enabled by the pseudocolor imaging method.

An iterative and targeted sampling design informed by habitat suitability models for detecting focal plant species over extensive areas

Prioritizing areas for management of non-native invasive plants is critical, as invasive plants can negatively impact plant community structure. Extensive and multi-jurisdictional inventories are essential to prioritize actions aimed at mitigating the impact of invasions and changes in disturbance regimes. However, previous work devoted little effort to devising sampling methods sufficient to assess the scope of multi-jurisdictional invasion over extensive areas. Here we describe a large-scale sampling design that used species occurrence data, habitat suitability models, and iterative and targeted sampling efforts to sample five species and satisfy two key management objectives: 1) detecting non-native invasive plants across previously unsampled gradients, and 2) characterizing the distribution of non-native invasive plants at landscape to regional scales. Habitat suitability models of five species were based on occurrence records and predictor variables derived from topography, precipitation, and remotely sensed data. We stratified and established field sampling locations according to predicted habitat suitability and phenological, substrate, and logistical constraints. Across previously unvisited areas, we detected at least one of our focal species on 77% of plots. In turn, we used detections from 2011 to improve habitat suitability models and sampling efforts in 2012, as well as additional spatial constraints to increase detections. These modifications resulted in a 96% detection rate at plots. The range of habitat suitability values that identified highly and less suitable habitats and their environmental conditions corresponded to field detections with mixed levels of agreement. Our study demonstrated that an iterative and targeted sampling framework can address sampling bias, reduce time costs, and increase detections. Other studies can extend the sampling framework to develop methods in other ecosystems to provide detection data. The sampling methods implemented here provide a meaningful tool when understanding the potential distribution and habitat of species over multi-jurisdictional and extensive areas is needed for achieving management objectives.

Using habitat suitability models to target invasive plant species surveys

Managers need new tools for detecting the movement and spread of nonnative, invasive species. Habitat suitability models are a popular tool for mapping the potential distribution of current invaders, but the ability of these models to prioritize monitoring efforts has not been tested in the field. We tested the utility of an iterative sampling design (i.e., models based on field observations used to guide subsequent field data collection to improve the model), hypothesizing that model performance would increase when new data were gathered from targeted sampling using criteria based on the initial model results. We also tested the ability of habitat suitability models to predict the spread of invasive species, hypothesizing that models would accurately predict occurrences in the field, and that the use of targeted sampling would detect more species with less sampling effort than a nontargeted approach. We tested these hypotheses on two species at the state scale (Centaurea stoebe and Pastinaca sativa) in Wisconsin (USA), and one genus at the regional scale (Tamarix) in the western United States. These initial data were merged with environmental data at 30-m2 resolution for Wisconsin and 1-km2 resolution for the western United States to produce our first iteration models. We stratified these initial models to target field sampling and compared our models and success at detecting our species of interest to other surveys being conducted during the same field season (i.e., nontargeted sampling). Although more data did not always improve our models based on correct classification rate (CCR), sensitivity, specificity, kappa, or area under the curve (AUC), our models generated from targeted sampling data always performed better than models generated from nontargeted data. For Wisconsin species, the model described actual locations in the field fairly well (kappa = 0.51, 0.19, P < 0.01), and targeted sampling did detect more species than nontargeted sampling with less sampling effort (chi2 = 47.42, P < 0.01). From these findings, we conclude that habitat suitability models can be highly useful tools for guiding invasive species monitoring, and we support the use of an iterative sampling design for guiding such efforts.

Geographical ecology of the palms (Arecaceae): determinants of diversity and distributions across spatial scales

BACKGROUND

The palm family occurs in all tropical and sub-tropical regions of the world. Palms are of high ecological and economical importance, and display complex spatial patterns of species distributions and diversity.

SCOPE

This review summarizes empirical evidence for factors that determine palm species distributions, community composition and species richness such as the abiotic environment (climate, soil chemistry, hydrology and topography), the biotic environment (vegetation structure and species interactions) and dispersal. The importance of contemporary vs. historical impacts of these factors and the scale at which they function is discussed. Finally a hierarchical scale framework is developed to guide predictor selection for future studies.

CONCLUSIONS

Determinants of palm distributions, composition and richness vary with spatial scale. For species distributions, climate appears to be important at landscape and broader scales, soil, topography and vegetation at landscape and local scales, hydrology at local scales, and dispersal at all scales. For community composition, soil appears important at regional and finer scales, hydrology, topography and vegetation at landscape and local scales, and dispersal again at all scales. For species richness, climate and dispersal appear to be important at continental to global scales, soil at landscape and broader scales, and topography at landscape and finer scales. Some scale-predictor combinations have not been studied or deserve further attention, e.g. climate on regional to finer scales, and hydrology and topography on landscape and broader scales. The importance of biotic interactions - apart from general vegetation structure effects - for the geographic ecology of palms is generally underexplored. Future studies should target scale-predictor combinations and geographic domains not studied yet. To avoid biased inference, one should ideally include at least all predictors previously found important at the spatial scale of investigation.

Remote sensing of plant functional types

Conceptually, plant functional types represent a classification scheme between species and broad vegetation types. Historically, these were based on physiological, structural and/or phenological properties, whereas recently, they have reflected plant responses to resources or environmental conditions. Often, an underlying assumption, based on an economic analogy, is that the functional role of vegetation can be identified by linked sets of morphological and physiological traits constrained by resources, based on the hypothesis of functional convergence. Using these concepts, ecologists have defined a variety of functional traits that are often context dependent, and the diversity of proposed traits demonstrates the lack of agreement on universal categories. Historically, remotely sensed data have been interpreted in ways that parallel these observations, often focused on the categorization of vegetation into discrete types, often dependent on the sampling scale. At the same time, current thinking in both ecology and remote sensing has moved towards viewing vegetation as a continuum rather than as discrete classes. The capabilities of new remote sensing instruments have led us to propose a new concept of optically distinguishable functional types ('optical types') as a unique way to address the scale dependence of this problem. This would ensure more direct relationships between ecological information and remote sensing observations.

Applications of remote sensing to alien invasive plant studies

Biological invasions can affect ecosystems across a wide spectrum of bioclimatic conditions. Therefore, it is often important to systematically monitor the spread of species over a broad region. Remote sensing has been an important tool for large-scale ecological studies in the past three decades, but it was not commonly used to study alien invasive plants until the mid 1990s. We synthesize previous research efforts on remote sensing of invasive plants from spatial, temporal and spectral perspectives. We also highlight a recently developed state-of-the-art image fusion technique that integrates passive and active energies concurrently collected by an imaging spectrometer and a scanning-waveform light detection and ranging (LiDAR) system, respectively. This approach provides a means to detect the structure and functional properties of invasive plants of different canopy levels. Finally, we summarize regional studies of biological invasions using remote sensing, discuss the limitations of remote sensing approaches, and highlight current research needs and future directions.