Spatial Analysis For Conservation and Sustainability
Remote Sensing
Satellite images provide a wonderful record of the last fifty years of global change. We have pioneered new methods to map wildlife habitat and proxies for biodiversity and habitat, as well as agricultural abandonment and other types of land use change for large areas. We analyze MODIS/VIIRS data across the globe, Landsat and Sentinel-2 across continents, and high-resolution CORONA spy satellite imagery across countries.
Land use change is a principal force and inherent element of global environmental change, threatening biodiversity, natural ecosystems, and their services. However, our ability to anticipate future land use change is severely limited by a lack of understanding of how major socio-economic disturbances (e.g., wars, revolutions, policy changes, and economic crises) affect land use. Here we explored to what extent socio-economic disturbances can shift land use systems onto a different trajectory, and whether this can result in less intensive land use. Our results show that the collapse of the Soviet Union in 1991 caused a major reorganization in land use systems. The effects of this socio-economic disturbance were at least as drastic as those of the nuclear disaster in the Chernobyl region in 1986. While the magnitudes of land abandonment were similar in Ukraine and Belarus in the case of the nuclear disaster (28% and 36% of previously farmed land, respectively), the rates of land abandonment after the collapse of the Soviet Union in Ukraine were twice as high as those in Belarus. This highlights that national policies and institutions play an important role in mediating effects of socio-economic disturbances. The socio-economic disturbance that we studied caused major hardship for local populations, yet also presents opportunities for conservation, as natural ecosystems are recovering on large areas of former farmland. Our results illustrate the potential of socio-economic disturbances to revert land use intensi?cation and the important role institutions and policies play in determining land use systems' resilience against such socio-economic disturbances.
Land use is a critical factor in the global carbon cycle, but land-use effects on carbon fluxes are poorly understood in many regions. One such region is Eastern Europe and the former Soviet Union, where land-use intensity decreased substantially after the collapse of socialism, and farmland abandonment and forest expansion have been widespread. Our goal was to examine how land-use trends affected net carbon ?uxes in western Ukraine (57 000 km2) and to assess the region's future carbon sequestration potential. Using satellite-based forest disturbance and farmland abandon- ment rates from 1988 to 2007, historic forest resource statistics, and a carbon bookkeeping model, we reconstructed carbon ?uxes from land use in the 20th century and assessed potential future carbon ?uxes until 2100 for a range of forest expansion and logging scenarios. Our results suggested that the low-point in forest cover occurred in the 1920s. Forest expansion between 1930 and 1970 turned the region from a carbon source to a sink, despite intensive logging during socialism. The collapse of the Soviet Union created a vast, but currently largely untapped carbon sequestration potential (up to = 150 Tg C in our study region). Future forest expansion will likely maintain or even increase the region's current sink strength of 1.48 Tg C yr-1. This may offer substantial opportunities for offsetting industrial carbon emissions and for rural development in regions with otherwise diminishing income opportunities. Through- out Eastern Europe and the former Soviet Union, millions of hectares of farmland were abandoned after the collapse of socialism; thus similar reforestation opportunities may exist in other parts of this region.
Conservation of biodiversity requires information at many spatial scales in order to detect and preserve habitat for many species, often simultaneously. Vegetation structure information is particularly important for avian habitat models and has largely been unavailable for large areas at the desired resolution. Airborne LiDAR, with its combination of relatively broad coverage and ?ne resolution provides existing new opportunities to map vegetation structure and hence avian habitat. Our goal was to model the richness of forest songbirds using forest structure information obtained from LiDAR data. In deciduous forests of southern Wisconsin, USA, we used discrete-return airborne LiDAR to derive forest structure metrics related to the height and density of vegetation returns, as well as composite variables that captured major forest structural elements. We conducted point counts to determine total forest songbird richness and the richness of foraging, nesting, and forest edge-related habitat guilds. A suite of 35 LiDAR variables were used to model bird species richness using best-subsets regression and we used hierarchical partitioning analysis to quantify the explanatory power of each variable in the multivariate models. Songbird species richness was correlated most strongly with LiDAR variables related to canopy and midstory height and midstory density (R 2 =0.204, pb0.001). Richness of species that nest in the midstory was best explained by canopy height variables (R 2 =0.197, pb0.001). Species that forage on the ground responded to mean canopy height and the height of the lower canopy (R 2 =0.149, pb0.005) while aerial foragers had higher richness where the canopy was tall and dense and the midstory more sparse (R 2 =0.216, pb0.001). Richness of edge-preferring species was greater where there were fewer vegetation returns but higher density in the understory (R 2 =0.153, pb0.005). Forest interior specialists responded positively to a tall canopy, developed midstory, and a higher proportion of vegetation returns (R 2 =0.195, pb0.001). LiDAR forest structure metrics explained between 15 and 20% of the variability in richness within deciduous forest songbird communities. This variability was associated with vertical structure alone and shows how LiDAR can provide a source of complementary predictive data that can be incorporated in models of wildlife habitat associations across broad geographical extents.
After the collapse of the Soviet Union, the forestry sector in Russia underwent substantial changes: the state forestry sector was decentralized, the timber industry was privatized, and timber use rights were allocated through short- and long-term leases. To date, there has been no quantitative assessment of the drivers of timber harvesting in European Russia following these changes. In this paper we estimate an econometric model of timber harvesting using remote sensing estimations of forest disturbance from 1990-2000 to 2000-2005 as our dependent variable. We aggregate forest disturbance to administrative districts - equivalent to counties in the United States - and test the impact of several biophysical and economic factors on timber harvesting. Additionally, we examine the impact that regions - equivalent to states in the United States and the main level of decentralized governance in Russia - have on timber harvesting by estimating the influence of regional-level effects on forest disturbance in our econometric model. Russian regions diverged considerably in political and economic conditions after the collapse of the Soviet Union, and the question is if these variations impacted timber harvesting after controlling for district-level biophysical and economic drivers. We find that the most important drivers of timber harvesting at the district level are road density, the percent of evergreen forest, and the total area of forest. The influence of these variables on timber harvesting changed over time and there was more harvesting closer to urban areas in 2000-2005. Even though district-level variables explain more than 70 percent of the variation in forest disturbance in our econometric model, we find that regional-level effects remain statistically significant. While we cannot identify the exact mechanism through which regional-level effects impact timber harvesting, our results suggest that sub-national differences can have a large and statistically significant impact on land-use outcomes and should be considered in policy design and evaluation.
Heat waves are expected to become more frequent and severe as climate changes, with unknown consequences for biodiversity. We sought to identify ecologically-relevant broad-scale indicators of heat waves based on MODIS land surface temperature (LST) and interpolated air temperature data and assess their associations with avian community structure. Speci ? cally, we asked which data source, time periods, and heat wave indices best predicted changes in avian abundance and species richness. Using mixed effects models, we analyzed associations between these indices and data from the North American Breeding Bird Survey in the central United States between 2000 and 2007 in four ecoregions and ? ve migratory and nesting species groups. We then quanti?ed avian responses to scenarios of severe, but commonly-occurring early, late, and summer-long heat waves. Indices based on MODIS LST data, rather than interpolated air temperatures, were more predictive of avian community structure. Avian communities were more related to 8-day LST exceedances (positive anomalies only); and were generally more sensitive to summer-long heat waves. Across the region, abundance, and to a lesser extent, species richness, declined following heat waves. Among the ecoregions, relationships were most consistently negative in the southern and montane ecoregions, but were positive in a more humid northern ecoregion. Among migratory groups, permanent resident species were the most sensitive, declining in abundance following a summer-long heat wave by 19% and 13% in the montane and southern ecoregions, respectively. Ground-nesting species, which declined in the south by 12% following a late summer heat wave, were more sensitive than avifauna overall. These results demonstrate the value of MODIS LST data for measuring ecologically-relevant heat waves across large regions. Ecologically, these ? ndings highlight the importance of extreme events for avian biodiversity and the considerable variation in response to environmental change associated with different functional groups and geographic regions. The magnitude of the relationships between avian abundance and heat waves reported here raises concerns about the impacts of more frequent and severe heat waves in a warming climate.
Farmland abandonment restructures rural landscapes in many regions worldwide in response to gradual industrialization and urbanization. In contrast, the political breakdown in Eastern Europe and the former Soviet Union triggered rapid and widespread farmland abandonment, but the spatial patterns of abandonment and its drivers are not well understood. Our goal was to map post-Socialist farmland abandonment in Western Ukraine using Landsat images from 1986 to 2008, and to identify spatial determinants of abandonment using a combination of best-subsets linear regression models and hierarchical partitioning. Our results suggest that farmland abandonment was widespread in the study region, with abandonment rates of up to 56%. In total, 6600 km 2 (30%) of the farmland used during socialism was abandoned after 1991. Topography, soil type, and population variables were the most important predictors to explain substantial spatial variation in abandonment rates. However, many of our a priori hypotheses about the direction of variable in?uence were rejected. Most importantly, abandonment rates were higher in the plains and lower in marginal areas. The growing importance of subsistence farming in the transition period, as well as off-farm income and remittances likely explain these patterns. The breakdown of socialism appears to have resulted in fundamentally different abandonment patterns in the Western Ukraine, where abandonment was a result of the institutional and economic shock, compared to those in Europe's West, where abandonment resulted from long-term socio-economic transformation such as urbanization and industrialization.
Niwaeli’s interest in smallholder woodlots started when she was conducting her Master’s research in Tanzania, at the southern end of the East African Rift. She saw rural farmers planting pine and eucalyptus in farms that are located near a forest edge on land that had been fallow for 15 years. Her collaborators, working at another forest edge site in the Uganda portion of the Rift, were noticing a similar conversion from cropland to tree plantations. She wanted to know how much smallholder tree planting was occurring near Rift forests, and why.
Niwaeli used Google Earth to get a first impression of the distribution of woodlots. She randomly selected 60 locations in a small portion of the Rift, and digitized all the woodlots in the area. So far the team has digitized more than 4000 individual woodlots. “The really striking thing is how small these woodlots are: 93% are less than 1 hectare, and the average area is 0.45 hectares”, she said. Niwaeli extrapolated the area of woodlots she found in the digitized subset to the entire Tanzanian Southern Highlands area and estimated 80.000 ha of smallholder woodlots. That does not sound like a big number, however, it is only ~ 10.000 ha less than the amount held by the biggest tree plantation owner, the Tanzanian government. A big question remains though: Why would a subsistence farmer plant trees with a slow turnaround rate, instead of growing food crops?
A Google Earth view showing woodlot patches (orange outlines) mixed with other land uses away from the forest edge.
In contrast to the US where farmers tend to own large and contiguous fields, in Niwaeli’s study site, farmers tend to own multiple small parcels of land that are spread throughout the landscape. Niwaeli thinks that the farmers make some allocation choices about which pieces of land will have crops, and which ones will have trees. She assumes that those parcels that are closer to the edge of the forests are not that great for growing food crops and are used for planting trees if the farmer can grow food somewhere else. “This could explain our initial field observations where we saw trees near forest edges” she added.
From the digitization, however, Niwaeli has seen that the woodlots are not limited to just forest edges, so the question of how farmers allocate land to trees is yet to be fully answered. She plans to further investigate the spatial distribution of tree plantations by mapping woodlots using Sentinel-2 satellite imagery. This will also provide a more accurate quantification of woodlot extent; and allow her to explore why some areas end up with woodlots while others do not.
Understanding past and current patterns of species richness is essential for predicting how these patterns may be affected by future global change. The species energy hypothesis predicts that higher abundance and richness of animal species occur where available energy is higher and more consistently available. There is a wide range of remote sensing proxies for available energy, such as vegetation productivity, but it is not clear which best predict species richness. Our goal here was to evaluate different proxies for annual plant productivity from Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) as input for the Dynamic Habitat Indices (DHIs), and to determine how well they predict the richness of breeding bird species in six functional guilds across the conterminous United States. The DHIs are measures of vegetation productivity over the course of a year and consist of three components: (1) cumulative productivity (DHI Cum), (2) minimum productivity (DHI Min), and (3) intra-annual variation of productivity (DHI Var). We hypothesized that increases in cumulative and minimum productivity and reductions in intra-annual variation will be associated with higher species richness. We calculated the DHIs from a range of MODIS 1000-m vegetation productivity data sets for 2003– 2014, i.e., the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Fraction of absorbed Photosynthetically Active Radiation (FPAR), Leaf Area Index (LAI), and Gross Primary Productivity (GPP). We summarized bird species richness of different guilds within ecoregions (n = 85) based on abundance maps derived from the N3000 routes of the North American Breeding Bird Survey for 2006 to 2012. Generally, we found all the DHIs had high explanatory power for predicting breeding bird species richness. However, the strength of the associations between the DHIs and bird species richness depended on habitat, nest placement, and migratory behavior. We found highest correlations for habitat-based guilds, such as grassland breeding species (R2 adj 0.66–0.73 for the multiple DHI regression model; R2 adj 0.41–0.61 for minimum DHI) and woodland breeding species (R2 adj 0.34–0.60 for the multiple DHI regression model; R2 adj 0.26–0.51 for cumulative DHI). The strong relationship between the DHIs and bird species richness reinforces the importance of vegetation productivity as a determinant of species diversity patterns, and the usefulness of satellite data for applying the species energy hypothesis to predictions in service to conservation.
Agricultural land abandonment is a common land-use change, making the accurate mapping of both location and timing when agricultural land abandonment occurred important to understand its environmental and social outcomes. However, it is challenging to distinguish agricultural abandonment from transitional classes such as fallow land at high spatial resolutions due to the complexity of change process. To date, no robust approach exists to detect when agricultural land abandonment occurred based on 30-m Landsat images. Our goal here was to develop a new approach to detect the extent and the exact timing of agricultural land abandonment using spatial and temporal segments derived from Landsat time series. We tested our approach for one Landsat footprint in the Caucasus, covering parts of Russia and Georgia, where agricultural land abandonment is widespread. First, we generated agricultural land image objects from multi-date Landsat imagery using a multiresolution segmentation approach. Second, we estimated the probability for each object that agricultural land was used each year based on Landsat temporal-spectral metrics and a random forest model. Third, we applied temporal segmentation of the resulting agricultural land probability time series to identify change classes and detect when abandonment occurred. We found that our approach was able to accurately separate agricultural abandonment from active agricultural lands, fallow land, and re-cultivation. Our spatial and temporal segmentation approach captured the changes at the object level well (overall mapping accuracy = 97 ± 1%), and performed substantially better than pixel-level change detection (overall accuracy = 82 ± 3%). We found strong spatial and temporal variations in agricultural land abandonment rates in our study area, likely a consequence of regional wars after the collapse of the Soviet Union. In summary, the combination of spatial and temporal segmentation approaches of time-series is a robust method to track agricultural land abandonment and may be relevant for other land-use changes as well.
People enjoy building houses in beautiful places where they are surrounded by the beauty of nature. Unfortunately, when wildfires rage through forests, these homes are often caught in the fire’s path. As more and more people attempt to enjoy the amenities of building a home in sparsely populated areas, communities increasingly face tough decisions whether to pay for protecting these homes from wildfires that destroy property and take lives. Wildfire costs are not trivial. During the twelve-years from 1999-2011, an average of 1,354 houses were destroyed and approximately $2 billion was spent fighting wildfires, annually. Ideally, communities would have information to help predict how wildfires spread and how to minimize the number of houses lost during wildfires. Unfortunately, a lot of basic information about what happens to a community after a wildfire rips though it is unknown. Patricia Alexandre and her colleagues recently published a study that makes a first step towards describing what happens to communities across the country following wildfire events. While their results suggest that the conventional wisdom that rebuilding always happens has little support and how much is rebuilt varies across the country, they were surprised to find that new housing constructed in burned areas was happening at higher rates than rebuilding, and often at higher rates than in surrounding non-fire areas, adding complexity to the discussion.
Example of a rebuilt building situation when using Google Earth
To be clear, Alexandre’s research is not intended to answer whether people should rebuild following a wildfire, but to provide a snapshot of what the patterns were within affected communities across the country following recent fires. This is an important step to take to see whether patterns are consistent across a large scale and provides a dataset to begin drawing conclusions from observed rebuilding patterns. To do this work, Alexandre refined a method utilizing historical images available on Google Earth and then recruited help from students to go through and hand-digitize structures present before a fire as well as all structures that had burned, and consequently been rebuilt within five years following a fire. Nationally, the team found rebuilding rates averaged 25%, with much higher rates in the western states. For example, rates in California approached 70% of structures rebuilt following a fire. The surprising results from Alexandre’s work is first that not all burned communities are re-building within five years following a fire, and second that new buildings were constructed in burned areas at similar or even higher rates. These results indicate that communities are not just replacing homes lost to wildfires, but many are putting new homes into burned areas.
Percentage of burned buildings per fire (fires occurred between 2000 and 2005)
Alexandre offers multiple reasons that homeowners may build, and rebuild, in burned areas, which are inherently fire prone. One reason is that homeowners may find the value they get from living in fire-prone areas worth the fire risk, some insurance policies require rebuilding in the same spot following a fire, and many homeowners do not have the finances to relocate to an area with lower fire risk. While the reasons to build and rebuild in fire prone areas likely vary widely across the country, Alexandre’s research provides a valuable baseline to evaluate future policies or practices that communities might use to mitigate wildfire damages. Whether it’s mandating that new or rebuilt structures be constructed with safer materials, or prohibiting rebuilding in burned areas, the best way to evaluate the efficacy of these policies is to compare them to the housing patterns before and after fire events. Alexandre’s research allows that comparison to take place and hopefully inform local initiatives that could save property, money and lives.”
