Animals select habitat at multiple spatial scales, suggesting that biodiversity modeling, for example of species richness, should be based on environmental data gathered at multiple spatial scales, and especially multiple grain sizes. Different satellite sensors collect data at different spatial resolutions and therefore provide opportunities for multi-grain habitat measures. The dynamic habitat indices (DHIs), which are derived from satellite data, capture patterns of vegetative productivity and predict bird species richness well. However, the DHIs have only been analyzed at single resolutions (e.g., 1-km), and have not yet been derived from high-resolution satellite data (< 10 -m). Our goal was to predict bird species richness based on measures of vegetation productivity (DHIs, NDVI median and NDVI percentile 90th) across a range of spatial resolutions both from different sensors, and from resampled high-resolution imagery. We analyzed bird species richness within 215 forest, grassland and shrubland plots (56.25 ha) located at 26 terrestrial field sites of the National Ecology Observatory Network (NEON), in the continental US. To obtain our multi-resolution measures of vegetation productivity, we acquired data from Planetscope (3-m), RapidEye (5-m), Sentinel-2 (10-m), Landsat-8 (30-m) and MODIS (250-m) from 2017 to 2020, generated time series of NDVI, calculated the three DHIs (cumulative, minimum and variation), NDVI median and the 90th percentile NDVI and calculated 1st and 2nd order texture measures. We evaluated the performance of the derived measures to predict bird species richness of habitat specialist guilds based on (i) univariate models (ii) multivariate models with single-resolution measures and (iii) multivariate models with multi-resolution measures. Single-spatial resolution measures predicted bird species richness moderately well (R2 up to 0.51) and the best performing spatial resolution and measure differed among bird species guilds. Highspatial resolution (3–5 m) measures outperformed medium-resolution measures (10–250 m). Models for all guilds performed best when incorporating multiple resolutions, including for all species richness (R2 = 0.63) and for forest (R2 = 0.72), grassland (R2 = 0.53) and shrubland specialists (R2 = 0.46). In addition, models based on multi-resolution data from different sensors performed better than models based on resampled high-resolution data for any of the guilds. Our results highlight, first, the value of the DHIs derived from high-resolution satellite data to predict bird species richness and, second, that remotely-sensed vegetation productivity measures from multiple spatial resolutions offer great promise for quantifying biodiversity.
File: 1-s2.0-S0034425723002122-main.pdf
Aim: Predicting biodiversity responses to global changes requires good models of
species' distributions. Both environmental conditions and human activities determine
population density patterns. However, quantifying the relationship between
wildlife population densities and their underlying environmental conditions across
large geographical scales has remained challenging. Our goal was to explain the
abundances of mammal species based on their response to several remotely sensed
indices including the Dynamic Habitat Indices (DHIs) and the novel Winter Habitat
Indices (WHIs).
Location: Russia, the majority of regions.
Taxon: Eight mammal species.
Methods: We estimated average population densities for each species across Russia
from 1981 to 2010 from winter track counts. The DHIs measure vegetative productivity,
a proxy for food availability. Our WHIs included the duration of snow-free
ground,
duration of snow-covered
ground and the start, end and length of frozen season. In
models, we included elevation, climate conditions, human footprint index. We parameterized
multiple linear regression and applied best-subset
model selection to determine
the main factors influencing population density.
Results: The DHIs were included in some of the top-twelve
models of every species,
and in the top model for moose, wild boar, red fox and wolf, so they were important
for species at all trophic levels. The WHIs were included in top models for all species
except roe deer, demonstrating the importance of winter conditions. The duration of
frozen ground without snow and the end of frozen season were particularly important.
Our top models performed well for all the species (R2
adj 0.43–0.87).
Main Conclusions: The combination of the DHIs and the WHIs with climate and
human-related
variables resulted in high explanatory power. We show that vegetation
productivity and winter conditions are key drivers of variation in population density
of eight species across Russia.
File: Razenkova-et-al-Abundance-patterns-of-mammals-across-Russia-explained-by-remotely-sensed.pdf
Global biodiversity loss is most pronounced in the tropics. Monitoring of broad-scale patterns of habitat is essential for biodiversity conservation. Image texture measures derived from satellite data are proxies for habitat heterogeneity, but have not been tested in tropical forests. Our goal was to evaluate image texture to predict tropical forest bird distributions across Thailand for different guilds. We calculated a suite of texture measures from cumulative productivity (1-km fPAR-MODIS data) for Thailand's forests, and assessed how well texture measures predicted distributions of 86 tropical forest bird species in relation to body size, and nesting guild. Finally, we compared the predictive performance of combining (a) satellite image texture measures, (b) habitat composition, and (c) habitat fragmentation. We found that texture measures predicted occurrences of tropical forest birds well (AUC = 0.801 ± 0.063). Second-order homogeneity was the most predictive texture measure. Our models based on texture were significantly better for birds with larger body size (p < 0.05), but did not differ among nesting guilds (p > 0.05). Models that combined texture with habitat composition measures (AUC = 0.928 ± 0.038) outperformed models that combined fragmentation with habitat composition measures (AUC = 0.905 ± 0.047) (p < 0.05). The incorporation of texture, composition, and fragmentation variables significantly improved model accuracy over texture-only models (AUC = 0.801 ± 0.063 to AUC = 0.938 ± 0.034; p < 0.05). We suggest that texture measures are a valuable tool to predict bird distributions at broad scales in tropical forests.
File: Suttidate-et-al_2023_Biological-Conservation.pdf
Fractional vegetation cover (FVC) is an essential input parameter for many environmental and ecological models. Recently, several global FVC products have been generated using remote sensing data. The Global LAnd Surface Satellite (GLASS) FVC product, which is generated from Moderate Resolution Imaging Spectroradiometer (MODIS) data, has attained acceptable performance. However, the original MODIS operation design lifespan has been exceeded. The Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-Orbiting Partnership (S-NPP) satellite was designed to be the MODIS successor. Therefore, developing an FVC estimation algorithm for VIIRS data is important for maintaining continuous FVC estimates in case of MODIS failure. In this study, a global FVC estimation algorithm for VIIRS surface reflectance data was proposed based on machine learning methods, which investigated the performances of back propagating neural networks (BPNNs), general regression networks (GRNNs), multivariate adaptive regression splines (MARS), and Gaussian process regression (GPR). The training samples were extracted from the GLASS FVC product and corresponding reconstructed VIIRS surface reflectance in 2013 over the global sampling locations. The VIIRS reflectances of red and near infrared (NIR) bands were the input variables for these machine learning methods. The theoretical performances and independent validation results indicated that the four machine learning methods could achieve similar and reliable FVC estimates. Regarding the FVC estimation accuracy, the GPR method achieved the best performance (R2 = 0.9019, RMSE = 0.0887). The MARS method had the obvious advantage of computational efficiency. Furthermore, the FVC estimates achieved good spatial and temporal continuities. Therefore, the proposed FVC estimation algorithm for VIIRS data can potentially generate reliable global FVC data for related applications.
File: Liu_RS_2018.pdf
Historical land use strongly influences current landscapes and ecosystems making maps of historical land cover an important reference point. However, the earliest satellite-based land cover maps typically date back to the 1980s only, after 30-m Landsat data became available. Our goal was to develop a methodology to automatically map land cover for large areas using high-resolution panchromatic Corona spy satellite imagery for 1964. Specifically, we a) conducted a comprehensive analysis on the feature selection and parameter setting for largearea classification processes for 2.5-m historical panchromatic Corona imagery for a full suite of land cover classes, b) compared the pixel-based and object-oriented methods of classifying the land cover, and c) examined the benefits of adding a digital elevation model for the pixel-based and object-oriented land cover classifications. We mapped land cover in parts of the Caucasus Mountains (158,000 km2), a study area with great variability in land cover types and illumination conditions. The overall accuracies of our pixel-based and object-oriented land cover maps were 63.0 ± 5.0% and 67.3 ± 4.0%, respectively, showing that object-oriented classifications performed better for Corona satellite data. Incorporating the digital elevation model improved the overall accuracy to 75.3 ± 3.0% and 78.7 ± 2.5%, respectively. The digital elevation model was especially useful for differentiating forest and snow-and-ice from lakes in mountainous areas affected by cast shadows. Our results highlight the feasibility of accurately and automatically classifying land cover for large areas based on Corona spy satellite imagery for the 1960s. Such land cover maps predate the earliest 30-m Landsat land cover classifications by two decades, and those from high-resolution satellite imagery by four decades. As such, we demonstrate here that Corona imagery can make important contributions to global change science.
File: Rizayeva_Nita_Radeloff_RSE_2023.pdf
Aim: The aim was to derive global indices of winter conditions and examine their relationships with species richness patterns outside of the tropics. Location: All extratropical areas (>25° N and 25° S latitudes), excluding islands. Time period: 2000– 2018.Major taxa studied: Amphibians, birds and mammals. Methods: We mapped three global indices of winter conditions [number of days of frozen ground (length of frozen ground winter); snow cover variability; and lack of subnivium (below-snow refuge)] from satellite data, then used generalized additive models to examine their relationships with species richness patterns derived from range data. Results: Length of frozen ground winter was the strongest predictor of species rich-ness, with a consistent cross-taxonomic decline in species richness occurring beyond 3 months of winter. It also often outperformed other environmental predictors of species richness patterns commonly used in biodiversity studies, including climate variables, primary productivity and elevation. In areas with ≥3 months of winter conditions, all three winter indices explained much of the deviance in amphibian, mammal and resident bird species richness. Mammals exhibited a stronger relationship with snow cover variability and lack of subnivium than the other taxa. Species richness of fully migratory species of birds peaked at c. 5.5 months of winter, coinciding with low species richness of residents. Main conclusions: Our study demonstrates that winter structures latitudinal and elevational gradients of extratropical terrestrial species richness. In a rapidly warming world, tracking the seasonal dynamics of frozen ground and snow cover will be essential for predicting the consequences of climate change on species, communities and ecosystems. The indices of winter conditions we developed from satellite imagery provide an effective means of monitoring these dynamics into the future.
File: GudexCross-Global-Ecology-and-Biogeography-2022-Winter-conditions-structure-extratropical-patterns-of-species.pdf
Large sets of autocorrelated data are common in fields such as remote sensing and genomics. For example, remote sensing can produce maps of information for millions of pixels, and the information from nearby pixels will likely be spatially autocorrelated. Although there are well-established statistical methods for testing hypotheses using autocorrelated data, these methods become computationally impractical for large datasets.
•The method developed here makes it feasible to perform F -tests, likelihood ratio tests, and t -tests for large autocorrelated datasets. The method involves subsetting the dataset into partitions, analyzing each partition separately, and then combining the separate tests to give an overall test.
•The separate statistical tests on partitions are non-independent, because the points in different partitions are not independent. Therefore, combining separate analyses of partitions requires accounting for the non- independence of the test statistics among partitions.
•The methods can be applied to a wide range of data, including not only purely spatial data but also spatiotemporal data. For spatiotemporal data, it is possible to estimate coefficients from time-series models at different spatial locations and then analyze the spatial distribution of the estimates. The spatial analysis can be simplified by estimating spatial autocorrelation directly from the spatial autocorrelation among time series.
File: Ives_MethodsX_2022.pdf
With the rapid advances of data acquisition techniques, spatio-temporal data are becoming increasingly abundant in a diverse array of disciplines. Here, we develop spatio-temporal regression methodology for analyzing large amounts of spatially referenced data collected over time, motivated by environmental studies utilizing remotely sensed satellite data. In particular, we specify a semiparametric autoregressive model without the usual Gaussian assumption and devise a computationally scalable procedure that enables the regression analysis of large datasets. We estimate the model parameters by maximum pseudolikelihood and show that the computational complexity can be reduced from cubic to linear of the sample size. Asymptotic properties under suitable regularity conditions are further established that inform the computational procedure to be efficient and scalable. A simulation study is conducted to evaluate the finite-sample properties of the parameter estimation and statistical inference. We illustrate our methodology by a dataset with over 2.96 million observations of annual land surface temperature, and comparison with an existing state-of-the-art approach to spatio-temporal regression highlights the advantages of our method.
File: TingFungMa_JABES_2022.pdf
Hundreds of millions of hectares of cropland have been abandoned globally since 1950 due to demographic, economic, and environmental changes. This abandonment has been seen as an important opportunity for carbon sequestration and habitat restoration; yet those benefits depend on the persistence of abandonment, which is poorly known. Here, we track abandonment and recultivation at 11 sites across four continents using annual land-cover maps for 1987–2017. We find that abandonment is largely fleeting, lasting on average only 14.22 years (SD = 1.44). At most sites, we project that >50% of abandoned croplands will be recultivated within 30 years, precluding the accumulation of substantial amounts of carbon and biodiversity. Recultivation resulted in 30.84% less abandonment and 35.39% less carbon accumulated by 2017 than expected without recultivation. Unless policy-makers take steps to reduce recultivation or provide incentives for regeneration, abandonment will remain a missed opportunity to reduce biodiversity loss and climate change.
File: Crawford_SciAdv_2022.pdf
Detailed maps of forest structure attributes are crucial for sustainable forest management, conservation, and forest ecosystem science at the landscape level. Mapping the structure of broad heterogeneous forests is chal lenging, but the integration of extensive field inventory plots with wall-to-wall metrics derived from synthetic aperture radar (SAR) and optical remote sensing offers a potential solution. Our goal was to map forest structure attributes (diameter at breast height, basal area, mean height, dominant height, wood volume and canopy cover) at 30-m resolution across the diverse 463,000 km2 of native forests of Argentina based on SAR Sentinel-1, vegetation metrics from Sentinel-2 and geographic coordinates. We modelled the forest structure attributes based on the latest national forest inventory, generated uncertainty maps, quantified the contribution of the predictors, and compared our height predictions with those from GEDI (Global Ecosystem Dynamics tion) and GFCH (Global Forest Canopy Height). We analyzed 3788 forest inventory plots (1000 m2 each) from Argentina’s Second Native Forest Inventory (2015–2020) to develop predictive random forest regression models. From Sentinel-1, we included both VV (vertical transmitted and received) and VH (vertical transmitted and horizontal received) polarizations and calculated 1st and 2nd order textures within 3 ×3 pixels to match the size of the inventory plots. For Sentinel-2, we derived EVI (enhanced vegetation index), calculated DHIs (dynamic habitat indices (annual cumulative, minimum and variation) and the EVI median, then generated 1st and 2nd order textures within 3 ×3 pixels of these variables. Our models including metrics from Sentinel-1 and 2, plus latitude and longitude predicted forest structure attributes well with root mean square errors (RMSE) ranging from 23.8% to 70.3%. Mean and dominant height models had notably good performance presenting relatively low RMSE (24.5% and 23.8%, respectively). Metrics from VH polarization and longitude were overall the most important predictors, but optimal predictors differed among the different forest structure attributes. Height predictions (r =0.89 and 0.85) outperformed those from GEDI (r =0.81) and the GFCH (r =0.66), suggesting that SAR Sentinel-1, DHIs from Sentinel-2 plus geographic coordinates provide great opportunities to map multiple forest structure attributes for large areas. Based on our models, we generated spatially-explicit maps of multiple forest structure attributes as well as uncertainty maps at 30-m spatial resolution for all Argentina’s native forest areas in support of forest management and conservation planning across the country.
File: Silveira_RSE_2022.pdf
