Marmots from space: assessing population size and habitat use of a burrowing mammal using publicly available satellite images

Abstract

Social, burrowing mammals such as prairie dogs, ground squirrels or marmots are keystone species in grassland ecosystems. Grasslands have been converted into cropland or pastures globally, yet it remains virtually unknown how this has affected the biogeography of burrowing mammals, as efficient, broad‐scale survey methods are lacking. We aimed to test whether structures created by burrowing rodents can be reliably detected on publicly available, very‐high‐resolution satellite images, in order to assess rodent distribution and abundance. We identified burrows of Bobak marmot (Marmota bobak), a keystone burrowing steppe rodent, on 1300 randomly selected plots of 1 km diameter (78.53 ha) across the species’ range (~950 000 km²) in Kazakhstan and southern Russia using Google Earth and Bing images. We then used burrow occurrences and species distribution models to map marmot distribution. We assessed how marmot occurrence and density vary across land‐use types. We also combined satellite‐based burrow densities and ground‐survey data to derive a new population estimate for the species across Kazakhstan. We mapped a total of 7425 burrows from the satellite imagery. Field visits at a subsample of burrows suggested that burrow occurrence was detected reliably. Broad‐scale marmot distribution was mainly determined by summer rainfall, land use and elevation. Occurrence probability was highest on arable croplands, followed by abandoned croplands and grazed steppe. The current Bobak marmot population size for Kazakhstan was estimated at 6.1 (±2.4) million individuals. Our results demonstrate that publicly available, very‐high‐resolution images can be used to reliably map the distribution of burrowing mammals across large geographic scales. The observed and predicted distributions indicate that the Bobak's range has remained almost unchanged in Kazakhstan since the 1950s, despite several drastic episode of land‐use change. This suggests that burrowing mammals can be remarkably resilient to land‐use pressure, questioning prevailing narratives of population collapse in these species following agricultural expansion.

Cold War spy satellite images reveal long-term declines of a philopatric keystone species in response to cropland expansion

“…Burrows appear as bright spots in both historical and contemporary imagery due to the large amount of soil turned by the marmots when digging and tending to the burrow (figure 1). Burrow location validation with field visits suggested that no false negatives occurred [33]. False positives only occurred in recently abandoned colonies (where burrows are usually covered by darker vegetation than the surrounding areas), but these were extremely scarce in our study area [33].…”

“…Other potential explanations for the high densities in croplands include a correlation between the most suitable marmot habitat and the suitable conditions for 'agriculture, a process locally described as 'colonies absorbed by agriculture' [51]. Indeed, our modelling results suggested that loamy soils had higher burrow densities compared to clayey or stony soils, which are less favourable for agriculture [33] (electronic supplementary material, S4). Finally, we caution that our study could not differentiate between grazed and ungrazed steppes, both combined in our single 'grassland' class.…”

“…Soil texture accounted for marmot preference towards soils that are easy to dig in, but stable enough to maintain burrow structure. We used normalized difference vegetation index (NDVI) measures for the month of May, shortly after marmots emerged from hibernation, as a proxy for food availability [33]. For each plot, we calculated Euclidean distances to the nearest river because marmots burrow along higher river banks and avoid areas with near-surface ground water [23].…”

“…For each plot, we calculated Euclidean distances to the nearest river because marmots burrow along higher river banks and avoid areas with near-surface ground water [23]. The distance to the nearest farm or livestock concentration point accounted for potential grazing interactions with livestock [33]. The average plot slope was accounted for because marmots prefer flat areas, with wide views that are advantageous for detecting predators.…”

“…Time-variant predictors included land use and distance to farms (electronic supplementary material, S2). Climate data were fed into the model as mean values since the 1960s [33] because climate change was negligible in Kazakhstan for this time period, particularly regarding climate effects on crop yields [47]. We standardized and centred all variables to improve the model interpretability.…”