Variable importance and scale of influence across individual scottish wildcat hybrid habitat models

Modelling species distributions with remote sensing data: bridging disciplinary perspectives

Species distribution models (SDMs) often rely on abiotic variables as proxies for biotic relationships. This means that important biotic relationships may be missed, creating ambiguity in our understanding of the drivers of habitat use. These problems are especially relevant for populations of predators, as their habitat use is likely to be strongly influenced by the distribution of their prey. We investigated habitat use of a population of a top predator, bottlenose dolphins Tursiops truncatus, in Doubtful Sound, New Zealand, using generalised additive models, and compared the results of models with and without biotic predictor variables. We found that although habitat use by bottlenose dolphins was significantly correlated with abiotic variables that likely describe foraging areas, introduction of biotic variables describing potential prey almost doubled the deviance explained, from 19.8 to 39.1%. Biotic variables were the most important of the predictors used, and indicated that the dolphins showed a preference for areas with a high abundance of a reef fish, girdled wrasse Notolabrus cinctus. For the dolphins of Doubtful Sound, these results show the importance of prey distribution in driving habitat use. On a broader scale, our results indicate that making an effort to include true biotic descriptors in SDMs can improve model performance, resulting in better understanding of the drivers of distribution of marine predators.

Species distribution models are generic empirical techniques that have a number of applications. One of these applications is to determine which environmental conditions are most important for a species. The calculation of this variable importance depends on a number of assumptions, including that the observations that are used to estimate the models are independent of each other. Spatial autocorrelation, which is a common feature most environmental factors confounds this assumption. Besides, many species distribution models are trained using a number of explanatory variables that have different levels of spatial autocorrelation. In this study we quantified the effects of differences in spatial autocorrelation in explanatory variables and the type of species responses to environmental gradients on variable importance estimations in species distribution models. We simulated data for both environmental predictors and species, so that we were in control of the true contribution of every variable in the model and the importance that could be estimated after fitting the models. We found that spatial autocorrelation in the predictors inflated the variable importance estimates, but only when the response of species to the environmental gradients is linear. This inflation effect was larger when the environmental preferences of species coincided with the dominant environmental conditions in a study site. Additionally we find that unimodal responses to the predictors yield systematically a higher variable importance compared to linear responses. We conclude that the type of response to environmental conditions and the relative levels of spatial autocorrelation in the environmental variables cause most bias in relative variable importance estimations. In this way, this study helps to clarify in a systematic and controlled approach how to make proper inferences about variable importance in species distribution models.

Continuous-based predictors of habitat characteristics derived from satellite imagery are increasingly used in species distribution models (SDM). This is especially the case of Normalized Difference Vegetation Index (NDVI) which provides estimates of vegetation productivity and heterogeneity. However, when NDVI predictors are incorporated into SDM, synchrony between biological observations and image acquisition must be questionned. Due to seasonal variations of NDVI during the year, landscape patterns of habitats are revealed differently from one date to another leading to variations in models’ performance. In this paper, we investigated the influence of acquisition time period of NDVI to explain and predict bird community patterns over France. We examined if the NDVI acquisition period that best fit the bird data depends on the dominant land cover context. We also compared models based on single time period of NDVI with one model built from the Dynamic Habitat Index (DHI) components which summarize variations in vegetation phenology throughout the year from the fraction of radiation absorbed by the canopy (fPAR). Bird species richness was calculated as response variable for 759 plots of 4 km2 from the French Breeding Bird Survey. Bird specialists and generalists to habitat were considered. NDVI and DHI predictors were both derived from MODIS products. For NDVI, five time periods in 2010 were compared, from late winter to begin of autumn. A climate predictor was also used and Generalized Additive Models were fitted to explain and predict bird species richness. Results showed that NDVI-based proxies of dominant habitat identity and spatial heterogeneity explain more bird community patterns than DHI-based proxies of annual productivity and seasonnality. We also found that models’ performance was both time and context-dependent, varying according to the bird groups. In general, best time period of NDVI did not match with the acquisition period of bird data because in case of synchrony, differences in habitats are less pronounced. These findings suggest that the most powerful approach to estimate bird community patterns is the simplest one. It only requires NDVI predictors from a single appropriate time period, in addition to climate, which makes the approach very operational.

Urban areas are generally several degrees hotter than surrounding rural areas, which is referred to as the urban heat island effect. Understanding the spatial pattern of land surface temperature (LST) in cities is important for urban planning, heat mitigation, and air pollution studies. This study developed five models to compare how land cover classes, normalized difference vegetation index (NDVI), social factors, and the combinations of social factors with land cover classes and NDVI perform in estimating LST at the census block group level. We found that all three groups of variables contribute significantly to understanding the spatial heterogeneity in LST. Land cover and NDVI explained LST to a similar extent (approx. 70%). Social factors alone explained 53% of LST's variance. Adding social factors in the combined models added little to the explanatory power (4.0%) compared to using land cover classes or NDVI along as independent variables.

For several fish species, increased river discharge has been linked to behavioral changes, including movement towards particular habitats. From 2000 to 2003, we used radiotelemetry to monitor habitat use by 42 muskellunge Esox masquinongy within a 17‐km reach of the New River, Virginia; to evaluate habitat selection; and to determine whether habitat use and selection were affected by discharge. Water depth, occurrence of aquatic vegetation, substrate type, and distance to riverbanks within the study reach were mapped with hydroacoustic and Global Positioning System technologies. Real‐time discharge measurements from an upstream U.S. Geological Survey stream gauge were used to classify fish locations into those made during periods of reduced (<75 m3/s) or increased (>75 m3/s) discharge. Seasonal habitat selection models for the different discharge levels were created using logistic regression. Differences in logistic regression coefficients between models were tested with multivariate chi‐square tests and Bonferroni‐corrected pairwise comparisons. Meta‐analytic averaging of Pearson's correlation coefficients was used to determine the overall effect size of discharge on habitat use by muskellunge. Overall, muskellunge exhibited a positive selection for deeper habitats; however, discharge was found to significantly affect both habitat use and selection. Habitat use was consistent with the hypothesis that fish moved to shallower habitats (both summer and winter) located closer to riverbanks (winter only) during periods of increased discharge. In terms of habitat selection, selection strengths for variables such as water depth (summer and winter) and distance to shoreline (winter only) generally weakened during periods of increased discharge, indicating that habitat use became more proportionate to availability. Increased movement as a result of increased discharge has been theorized to potentially reduce winter survival of fish because of seasonal energetic limitations. Thus, indirect effects on fish from dam operating modes (e.g., hydropeaking) may be as important as direct effects when evaluating biotic implications of water release schedules.

The Hawaiian Islands have been identified as a global biodiversity hotspot. We examine the Normalized Difference Vegetation Index (NDVI) using Climate Data Records products (0.05 × 0.05°) to identify significant differences in NDVI between neutral El Niño-Southern Oscillation years (1984, 2019) and significant long-term changes over the entire time series (1982–2019) for the Hawaiian Islands and six land cover classes. Overall, there has been a significant decline in NDVI (i.e., browning) across the Hawaiian Islands from 1982 to 2019 with the islands of Lāna’i and Hawai’i experiencing the greatest decreases in NDVI (≥44%). All land cover classes significantly decreased in NDVI for most months, especially during the wet season month of March. Native vegetation cover across all islands also experienced significant declines in NDVI, with the leeward, southwestern side of the island of Hawai’i experiencing the greatest declines. The long-term trends in the annual total precipitation and annual mean Palmer Drought Severity Index (PDSI) for 1982–2019 on the Hawaiian Islands show significant concurrent declines. Primarily positive correlations between the native ecosystem NDVI and precipitation imply that significant decreases in precipitation may exacerbate the decrease in NDVI of native ecosystems. NDVI-PDSI correlations were primarily negative on the windward side of the islands and positive on the leeward sides, suggesting a higher sensitivity to drought for leeward native ecosystems. Multi-decadal time series and spatially explicit data for native landscapes provide natural resource managers with long-term trends and monthly changes associated with vegetation health and stability.

Mapping abandoned agriculture with multi-temporal MODIS satellite data

Accurate vegetation detection is important for many applications, such as crop yield estimation, land cover land use monitoring, urban growth monitoring, drought monitoring, etc. Popular conventional approaches to vegetation detection incorporate the normalized difference vegetation index (NDVI), which uses the red and near infrared (NIR) bands, and enhanced vegetation index (EVI), which uses red, NIR, and the blue bands. Although NDVI and EVI are efficient, their accuracies still have room for further improvement. In this paper, we propose a new approach to vegetation detection based on land cover classification. That is, we first perform an accurate classification of 15 or more land cover types. The land covers such as grass, shrub, and trees are then grouped into vegetation and other land cover types such as roads, buildings, etc. are grouped into non-vegetation. Similar to NDVI and EVI, only RGB and NIR bands are needed in our proposed approach. If Laser imaging, Detection, and Ranging (LiDAR) data are available, our approach can also incorporate LiDAR in the detection process. Results using a well-known dataset demonstrated that the proposed approach is feasible and achieves more accurate vegetation detection than both NDVI and EVI. In particular, a Support Vector Machine (SVM) approach performed 6% better than NDVI and 50% better than EVI in terms of overall accuracy (OA).

Ndvi: Vegetation Change Detection Using Remote Sensing and Gis – A Case Study of Vellore District

Climate change has a significant impact on vegetation development. While existing studies provide some insights, long-term trend analysis and multifactor driver assessments for China are still lacking. At the same time, research on the future vegetation development under different climate change scenarios needs further strengthening. In response to these issues, this study analyzed China’s normalized difference vegetation index (NDVI) data from 2001 to 2023, exploring vegetation cover trends, driving factors, and predicting the impact of future climate change. Firstly, this study decomposed the time series data into seasonal, trend, and residual components using the Seasonal–Trend decomposition using Loess (STL) decomposition method, quantifying vegetation changes across different climate zones. Partial least squares (PLS) regression analysis was then used to examine the relationship between NDVI and driving factors, and the contribution of these factors to NDVI variation was determined through the variable importance in projection (VIP) score. The results show that NDVI has significantly increased over the past two decades, especially since 2010. Further analysis revealed that vegetation growth is primarily influenced by soil moisture, shortwave radiation, and total precipitation (VIP scores > 0.8). Utilizing machine learning with Coupled Model Intercomparison Project Phase 6 (CMIP6) multimodel data, this study predicts NDVI trends from 2023 to 2100 under four emission scenarios (SSP126, SSP245, SSP370, SSP585), quantifying future meteorological factors such as temperature, precipitation, and radiation to NDVI. Findings indicate that under high-emission scenarios, the vegetation greenness in some regions may experience improved vegetation conditions despite global warming challenges. Future land management strategies must consider climate change impacts on ecosystems to ensure sustainability and enhance ecosystem services.

The present study explores vegetation health and forest canopy density in the Sundarbans region using Landsat-8 images. This work analyzes changes in vegetation health using two vegetation indices, the Normalized Difference Vegetation Index (NDVI) and Forest Canopy Density (FCD) values of the Sundarbans, from 2014 to 2020. NDVI, comprising two bands, Red and Near-infrared (NIR), shows a declining trend during the period. Two NDVI land cover classification maps for 2014 and 2020 are produced, and the interest area is divided into five classes: Scanty, Low, Medium, and Densely Vegetated Regions and Water Bodies. A single-band linear gradient pseudo-color is used to assess the land cover difference between 2020 and 2014, showing marked changes in densely vegetative areas. The NDVI difference marks the coastal regions with a higher depletion rate of vegetation than the regions away from the seacoasts. FCD has been taken to compare the results of NDVI with it. FCD consists of another four models: AVI (advanced vegetative index), BI (Bare soil index), SSI (scaled shadow index), and TI (thermal index). FCD is also called crown cover or canopy coverage, which refers to the portion of an area in the field covered by the crown of trees. 2014 and 2015 FCD maps are produced with a single band linear gradient pseudocolor with five land cover classifications: bare soil, Bare Soil, Shrubs, Low, Medium, and Highly vegetated regions. Both maps bear a significant resemblance to NDVI land classification maps. Further, the FCD values of the two maps are scaled between 1 and 100, and the area of each class is calculated. To check the veracity of the NDVI and FCD analysis, a Deep Neural Network (DNN) model has been developed to classify each year’s image taken from Google Earth Engine (GEE). It classifies each year’s image with 99% accuracy. The calculation of the area of each class emphasizes the rapid decline of densely wooded vegetation. Almost 80% of the highly forested zone has been diminished and has become part of the medium-forested region. Area inflation in medium-forested regions corroborates the same. The study also analyzes the migration of vegetation density, i.e., where and how many areas are unchanged, growing, or deforested.

In this study, the Savitzky-Golay filter was applied to smooth observed unnatural variations in the temporal profiles of the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI) time series from the MODerate Resolution Imaging Spectroradiometer (MODIS). We computed two sets of land cover classifications based on the NDVI and EVI time series before and after applying the Savitzky-Golay filter. The resulting classification from the filtered versions of the vegetation indices showed a substantial improvement in accuracy when compared to the classifications from the unfiltered versions. The classification by the EVIsg had the highest K (0.72) for all classes compared to those of the EVI (0.67), NDVI (0.63), and NDVIsg (0.62). Therefore, we conclude that the EVIsg is best suited for land cover classification compared to the other data sets in this study.

Temporal-related features are important for improving land cover classification accuracy using remote sensing data. This study investigated the efficacy of phenological features extracted from time series MODIS Normalized Difference Vegetation Index (NDVI) data in improving the land cover classification accuracy of Landsat data. The MODIS NDVI data were first fused with Landsat data via the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) algorithm to obtain NDVI data at the Landsat spatial resolution. Next, phenological features, including the beginning and ending dates of the growing season, the length of the growing season, seasonal amplitude, and the maximum fitted NDVI value, were extracted from the fused time series NDVI data using the TIMESAT tool. The extracted data were integrated with the spectral data of the Landsat data to improve classification accuracy using a maximum likelihood classifier (MLC) and support vector machine (SVM) classifier. The results indicated that phenological features had a statistically significant effect on improving the land cover classification accuracy of single Landsat data (an approximately 3% increase in overall classification accuracy), especially for vegetation type discrimination. However, the phenological features did not improve on statistical measures including the maximum, the minimum, the mean, and the standard deviation values of the time series NDVI dataset, especially for human-managed vegetation types. Regarding different classifiers, SVM could achieve better classification accuracy than the traditional MLC classifier, but the improvement in accuracy obtained using advanced classifiers was inferior to that achieved by involving the temporally derived features for land cover classification.

Abstract. Climate variables carry signatures of variability at multiple timescales. How these modes of variability are reflected in the state of the terrestrial biosphere is still not quantified or discussed at the global scale. Here, we set out to gain a global understanding of the relevance of different modes of variability in vegetation greenness and its covariability with climate. We used >30 years of remote sensing records of the normalized difference vegetation index (NDVI) to characterize biosphere variability across timescales from submonthly oscillations to decadal trends using discrete Fourier decomposition. Climate data of air temperature (Tair) and precipitation (Prec) were used to characterize atmosphere–biosphere covariability at each timescale. Our results show that short-term (intra-annual) and longer-term (interannual and longer) modes of variability make regionally highly important contributions to NDVI variability: short-term oscillations focus in the tropics where they shape 27 % of NDVI variability. Longer-term oscillations shape 9 % of NDVI variability, dominantly in semiarid shrublands. Assessing dominant timescales of vegetation–climate covariation, a natural surface classification emerges which captures patterns not represented by conventional classifications, especially in the tropics. Finally, we find that correlations between variables can differ and even invert signs across timescales. For southern Africa for example, correlation between NDVI and Tair is positive for the seasonal signal but negative for short-term and longer-term oscillations, indicating that both short- and long-term temperature anomalies can induce stress on vegetation dynamics. Such contrasting correlations between timescales exist for 15 % of vegetated areas for NDVI with Tair and 27 % with Prec, indicating global relevance of scale-specific climate sensitivities. Our analysis provides a detailed picture of vegetation–climate covariability globally, characterizing ecosystems by their intrinsic modes of temporal variability. We find that (i) correlations of NDVI with climate can differ between scales, (ii) nondominant subsignals in climate variables may dominate the biospheric response, and (iii) possible links may exist between short-term and longer-term scales. These heterogeneous ecosystem responses on different timescales may depend on climate zone and vegetation type, and they are to date not well understood and do not always correspond to transitions in dominant vegetation types. These scale dependencies can be a benchmark for vegetation model evaluation and for comparing remote sensing products.