Ecological regionalisations delineate areas of similar environmental conditions, ecological processes,
and biotic communities, and provide a basis for systematic conservation planning and management.
Most regionalisations are made based on subjective criteria, and can not be readily revised, leading to
outstanding questions with respect to how to optimally develop and define them. Advances in remote
sensing technology, and big data analysis approaches, provide new opportunities for regionalisations,
especially in terms of productivity patterns through both photosynthesis and structural surrogates.
Here we show that global terrestrial productivity dynamics can be captured by Dynamics Habitat
Indices (DHIs) and we conduct a regionalisation based on the DHIs using a two-stage multivariate
clustering approach. Encouragingly, the derived clusters are more homogeneous in terms of species
richness of three key taxa, and of canopy height, than a conventional regionalisation. We conclude
with discussing the benefits of these remotely derived clusters for biodiversity assessments and
conservation. The clusters based on the DHIs explained more variance, and greater within-region
homogeneity, compared to conventional regionalisations for species richness of both amphibians and
mammals, and were comparable in the case of birds. Structure as defined by global tree height was
also better defined by productivity driven clusters than conventional regionalisations. These results
suggest that ecological regionalisations based on remotely sensed metrics have clear advantages over
conventional regionalisations for certain applications, and they are also more easily updated.
File: Coop2018_clusters_Nature.pdf
This is a publication uploaded with a php script
Remotely sensed data can help to identify both suitable habitat for individual species, and environmental conditions
that foster species richness, which is important when predicting how biodiversity will respond to global change. The
question is how to summarize remotely sensed data so that they are most relevant for biodiversity analyses, and the
Dynamic Habitat Indices are three metrics designed for this. Our goals here were to a) derive, for the first time, the
Dynamic Habitat Indices (DHIs) globally, and b) use these to evaluate three hypotheses (available energy, environmental
stress, and environmental stability) that attempt to explain global variation in species richness of
amphibians, birds, and mammals. The three DHIs summarize three key measures of vegetative productivity: a)
annual cumulative productivity, which we used to evaluate the available energy hypothesis that more energy is
associate with higher species richness; b) minimum productivity throughout the year, which we used to evaluate the
environmental stress hypothesis that higher minima cause higher species richness, and c) seasonality, expressed as
the annual coefficient of variation in productivity, which we used to evaluate the environmental stability hypothesis
that less intra-annual variability causes higher species richness. We calculated the DHIs globally at 1-km resolution
from MODIS vegetation products (NDVI, EVI, LAI, fPAR, and GPP), based on the median of the good observations of
all years from the entire MODIS record for each of the 23 or 46 possible dates (8- vs. 16-day composites) during the
year, and calculated species richness for three taxa (amphibians, birds, and mammals) at 110-km resolution from
species range maps from the IUCN Red List. We found marked global patterns of the DHIs, and strong support for all
three hypotheses. The three DHIs for a given vegetation product were well correlated (Spearman rank correlations
ranging from −0.6 (cumulative vs. variation DHIs) to −0.93 (variation vs. minimum DHI)). Similarly, DHI components
derived from different MODIS vegetation products were well correlated (0.8–0.9), and correlations of the
DHIs with temperature and precipitation were moderate and strong respectively. All three DHIs were well correlated
with species richness, showing in ranked order positive correlations for cumulative DHI based on GPP (Spearman
rank correlations of 0.75, 0.63, and 0.67 for amphibians, resident birds, and mammals respectively) and minimum
DHI (0.73, 0.83, and 0.62), and negative for variation DHI (−0.69, −0.83, and −0.59). Multiple linear models of all
three DHIs explained 67%, 65%, and 61% of the variability in species richness of amphibians, resident birds, and
mammals, respectively. The DHIs, which are closely related to well-established ecological hypotheses of biodiversity,
can predict species richness well, and are promising for application in biodiversity science and conservation.
File: Radeloff_Global_DHIs_RSE_2019.pdf
Aim: Populations of large ungulates are dwindling worldwide. This is especially so for
wild sheep, which compete with livestock for forage, are disturbed by shepherds and
their dogs, and are exposed to disease transmissions from livestock. Our aim was to
assess spatial patterns in realized niche overlap between wild and domestic sheep to
better understand where potential competition might arise, and thus to identify priority
areas for wild sheep recovery.
Location: Southern Caucasus (220,000 km2).
Methods: We studied Gmelin’s mouflon (Ovis orientalis gmelinii), an ancestor of domestic
sheep, to investigate seasonal habitat use and niche overlap with domestic
sheep. To map habitat, we analysed mouflon occurrences collected during 2006–
2016, and domestic sheep occurrences from shepherd camp locations digitized on
high-resolution
satellite imagery. We mapped areas of potential competition between
mouflon and domestic sheep and assessed potential habitat displacement.
Results: Mouflon and domestic sheep niches overlapped substantially (overlap index
I = 0.89, where 1 means perfect overlap) but were not identical. Mouflon habitat was
less widespread than domestic sheep habitat (14,000 vs. 40,270 km2) and tended to
be located in more rugged areas with less vegetation cover. We identified 51 priority
patches as reintroduction candidates if grazing pressure and poaching were
reduced.
Main conclusions: Our results suggest that competition with domestic sheep might
have pushed mouflon into marginal habitat. Thus, conservation efforts focusing on
current mouflon habitat might miss suitable reintroduction sites. We demonstrate
that a combined habitat model for wild and domestic sheep can identify general
sheep habitat, which might be more useful for conservation planning than understanding
current mouflon habitat selection. Our results highlight that considering
competition with livestock is important for large ungulate conservation, both in
terms of reactive (e.g., lessening livestock pressure in prime habitat) and proactive
strategies (e.g., reintroduction in areas with low contemporary overlap).
File: Bleyhl_et_al-2019-Diversity_and_Distributions.pdf
This is a publication uploaded with a php script
Armed conflicts are globally widespread and can strongly influence societies and the environment. However,
where and how armed conflicts affect agricultural land-use is not well-understood. The Caucasus is a multiethnic
region that experienced several conflicts shortly after the collapse of the Soviet Union, most notably the
two Chechen Wars, raising the question how agricultural lands were changed. Here, we investigated how the
distance to conflicts and conflict intensity, measured as the number of conflicts and the number of casualties,
affected agricultural land abandonment and subsequent re-cultivation, by combining social, environmental and
economic variables with remotely-sensed maps of agricultural change. We applied logistic and panel regression
analyses for both the First Chechen War (1994–1996) and the Second Chechen War (1999–2009) and interacted
conflict distance with conflict intensity measures. We found that agricultural lands closer to conflicts were more
likely to be abandoned and less likely to be re-cultivated, with stronger effects for the First Chechen War.
Conflict intensity was positively correlated with agricultural land abandonment, but the effects differed based on
distance to conflicts and the intensity measure. We found little re-cultivation after the wars, despite abundant
subsidies, indicating the potentially long-lasting effects of armed conflicts on land-use. Overall, we found a clear
relationship between the Chechen Wars and agricultural land abandonment and re-cultivation, illustrating the
strong effects of armed conflicts on agriculture.
File: Yin_etal_GEC_2019.pdf
This is a publication uploaded with a php script
Refugia are habitats that allow organisms to persist when the environment makes persistence impossible elsewhere. The subnivium—the interface between snowpack and ground—is an important seasonal refugium that protects diverse species from extreme winter temperatures, but its future duration is uncertain with climate change. Here, we predict that subnivium duration will decrease from 126 d (2010–2014) to 110 d (2071–2100), which we have inferred using past and future duration of frozen ground with snow cover (Dsc) derived from remotely sensed datasets and climate projections. Concomitantly, duration of frozen ground without snow cover (Dfwos) at mid-latitudes is predicted to increase from 35 d to 45 d, with notable increases in the western United States, Europe, the Tibetan Plateau and Mongolia. In most areas, increasing winter temperatures were more important than precipitation for decreasing Dsc and increasing Dfwos. Thus, counter-intuitively, warming climate will cause longer Dfwos at mid-latitudes, causing functional winter cooling for subnivium-dependent organisms.
File: Zhu_etal_NatureArticles_2019.pdf
Refugia are habitats that allow organisms to persist when the environment makes persistence impossible elsewhere. The subnivium—the interface between snowpack and ground—is an important seasonal refugium that protects diverse species from extreme winter temperatures, but its future duration is uncertain with climate change. Here, we predict that subnivium duration will decrease from 126 d (2010–2014) to 110 d (2071–2100), which we have inferred using past and future duration of frozen ground with snow cover (Dsc) derived from remotely sensed datasets and climate projections. Concomitantly, duration of frozen ground without snow cover (Dfwos) at mid-latitudes is predicted to increase from 35 d to 45 d, with notable increases in the western United States, Europe, the Tibetan Plateau and Mongolia. In most areas, increasing winter temperatures were more important than precipitation for decreasing Dsc and increasing Dfwos. Thus, counter-intuitively, warming climate will cause longer Dfwos at mid-latitudes, causing functional winter cooling for subnivium-dependent organisms.
Aim: To evaluate how environment and evolutionary history interact to influence
global patterns of mammal trait diversity (a combination of 14 morphological and
life-history traits).
Location: The global terrestrial environment.
Taxon: Terrestrial mammals.
Methods: We calculated patterns of spatial turnover for mammalian traits and phylogenetic
lineages using the mean nearest taxon distance. We then used a variance
partitioning approach to establish the relative contribution of trait conservatism,
ecological adaptation and clade specific ecological preferences on global trait
turnover.
Results: We provide a global scale analysis of trait turnover across mammalian terrestrial
assemblages, which demonstrates that phylogenetic turnover by itself does
not predict trait turnover better than random expectations. Conversely, trait turnover
is consistently more strongly associated with environmental variation than
predicted by our null models. The influence of clade-specific ecological preferences,
reflected by the shared component of phylogenetic turnover and environmental
variation, was considerably higher than expectations. Although global patterns of
trait turnover are dependent on the trait under consideration, there is a consistent
association between trait turnover and environmental predictive variables, regardless
of the trait considered.
Main conclusions: Our results suggest that changes in phylogenetic composition
are not always coupled with changes in trait composition on a global scale and that
environmental conditions are strongly associated with patterns of trait composition
across species assemblages, both within and across phylogenetic clades.
File: Holt2018_jbi.13091.pdf
This is a publication uploaded with a php script
Global variation in species richness is widely recognized, but the explanation
for what drives it continues to be debated. Previous efforts have focused on a
subset of potential drivers, including evolutionary rate, evolutionary time
(maximum clade age of species restricted to a region), dispersal (migration
from one region to another), ecological factors and climatic stability. However,
no study has evaluated these competing hypotheses simultaneously at a broad
spatial scale. Here, we examine their relative contribution in determining the
richness of the most comprehensive dataset of tetrapods to our knowledge
(84% of the described species), distinguishing between the direct influences of
evolutionary rate, evolutionary time and dispersal, and the indirect influences
of ecological factors and climatic stability through their effect on direct factors.
We found that evolutionary time exerted a primary influence on species richness,
with evolutionary rate being of secondary importance. By contrast,
dispersal did not significantly affect richness patterns. Ecological and climatic
stability factors influenced species richness indirectly by modifying evolutionary
time (i.e. persistence time) and rate. Overall, our findings suggest that
global heterogeneity in tetrapod richness is explained primarily by the
length of time species have had to diversify.
File: Marin-et-al.-2018.pdf
This is a publication uploaded with a php script
Wars have major economic, political and human implications, and they can strongly affect environment and land
use, not only during the conflicts, but also afterwards. However, data on the land use effects of wars is sparse,
especially for World War II, the largest war in history. Our goal was to quantify and understand the time-lagged
land use effects of WWII in Romania, by applying Structure from Motion technology to 1960s Corona spy satellite
photography. We quantified forest cutting across Romania from 1955 to 1965. This was a period when
Romania's economy recovered from the war and when Romania established close economic ties to the Soviet
Union, and when the Romanian government made reparation payments to the Soviet Union. To understand the
effects of war, we developed an accurate and fast method to orthorectify high-resolution Corona photography in
mountain areas, and rectified scanned Corona photography based on Structure from Motion technology. Our
study area of 212,000 km2 was covered by 208 Corona film strips, which we rectified with an overall average
accuracy of 14.3 m. We identified 530,000 ha of forest cuts over this time period, the rate of which is three times
higher than contemporary cutting rates. Our results highlight that the environmental and land use effects of
WWII were substantial in Romania, due to reparation payments, post-war policies regarding resource exploitation,
and technological and infrastructural development. Our research provides quantitative evidence of
how wars can cause time-lagged and long-term effects on the environment. Methodologically, we advance remote
sensing science by pioneering a new approach to orthorectify Corona photography for large areas effectively.
Corona data are available globally. Our approach facilitates the extension of the data record of space
borne observation of the earth by one to two decades earlier than what is possible with satellite datasets.
File: Nita2018_WWII_RSE.pdf
This is a publication uploaded with a php script
Mapping crop types is of great importance for assessing agricultural production, land-use patterns, an the environmental effects of agriculture. Indeed, both radiometric and spatial resolution of Landsat’s sensor images are optimized for cropland monitoring. However, accurate mapping of crop types require frequent cloud-free images during the growing season, which are often not available, and this raises th question of whether Landsat data can be combined with data from other satellites. Here, our goal is t evaluate to what degree fusing Landsat with MODIS Nadir Bidirectional Reflectance Distribution Functio (BRDF)-Adjusted Reflectance (NBAR) data can improve crop-type classification. Choosing either one o two images from all cloud-free Landsat observations available for the Arlington Agricultural Researc Station area in Wisconsin from 2010 to 2014, we generated 87 combinations of images, and used eac combination as input into the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) algorith to predict Landsat-like images at the nominal dates of each 8-day MODIS NBAR product. Bot the original Landsat and STARFM-predicted images were then classified with a support vector machin (SVM), and we compared the classification errors of three scenarios: 1) classifying the one or two origina Landsat images of each combination only, 2) classifying the one or two original Landsat image plus all STARFM-predicted images, and 3) classifying the one or two original Landsat images togethe with STARFM-predicted images for key dates. Our results indicated that using two Landsat images as th input of STARFM did not significantly improve the STARFM predictions compared to using only one, an predictions using Landsat images between July and August as input were most accurate. Including al STARFM-predicted images together with the Landsat images significantly increased average classificatio error by 4% points (from 21% to 25%) compared to using only Landsat images. However, incorporatin only STARFM-predicted images for key dates decreased average classification error by 2% points (fro 21% to 19%) compared to using only Landsat images. In particular, if only a single Landsat image was available adding STARFM predictions for key dates significantly decreased the average classification error b 4 percentage points from 30% to 26% (p < 0.05). We conclude that adding STARFM-predicted images ca be effective for improving crop-type classification when only limited Landsat observations are available but carefully selecting images from a full set of STARFM predictions is crucial. We developed an approac to identify the optimal subsets of all STARFM predictions, which gives an alternative method of featur selection for future research.
File: Zhu2017_CropMODIS_IJAEOG.pdf
This is a publication uploaded with a php script
When timber harvesting is an important source of
local income and forest resources are declining, even
forests that are designated as protected areas may
become vulnerable. Therefore, regular monitoring
of forest disturbance is necessary to enforce the
protection of forest ecosystems. However, mapping
forest disturbance with satellite imagery can be
complicated if the majority of the harvesting is
selective logging and not clearcuts. Our goal was to
map both selective logging and clearcuts within and
outside of protected areas in Western Siberia, a region
with a highly developed timber industry. Combining
summer and winter imagery allowed us to accurately
estimate not only clearcuts, but also selective logging.
Winter Landsat images substantially improved our
classification and resulted in a highly accurate forest
disturbance map (97.5% overall accuracy and 86%
user accuracy for the rarest class, clearcuts). Selective
logging and stripcuts were the dominant disturbance
types, accounting for 96.3% of all forest disturbances,
versus 3.7% for clearcuts. The total annual forest
disturbance rate (i.e. disturbance rate for clearcuts,
stripcuts and selective logging together) was 0.53%,
but total forest disturbance within protected areas
was greater than in non-protected forest (0.66% versus
0.50%, respectively), and so was the annual rate
of selective logging (i.e. without clearcuts, 0.37%
versus 0.25%, respectively). Our results highlight that
monitoring only clearcuts without assessing selective
logging might result in significant underestimation of
forest disturbance. Also, when timber harvesting is
important for the local economy and when protected
areas have valuable timber resources that have already
been depleted elsewhere, then additional protection
may be necessary in order to maintain natural forests
within protected areas. We suggest that this is the
situation in our study area in Western Siberia right
now and is likely the situation in many other parts of
the globe as well.
File: Shchur2017_monitoring_selective_logging_with_landsat_EnvCons.pdf
This is a publication uploaded with a php script
