Elephant movement closely tracks precipitation-driven vegetation dynamics in a Kenyan forest-savanna landscape

Abstract

Background: This study investigates the ranging behavior of elephants in relation to precipitation-driven dynamics of vegetation. Movement data were acquired for five bachelors and five female family herds during three years in the Marsabit protected area in Kenya and changes in vegetation were mapped using MODIS normalized difference vegetation index time series (NDVI). In the study area, elevations of 650 to 1100 m.a.s.l experience two growth periods per year, while above 1100 m.a.s.l. growth periods last a year or longer.

Results: We find that elephants respond quickly to changes in forage and water availability, making migrations in response to both large and small rainfall events. The elevational migration of individual elephants closely matched the patterns of greening and senescing of vegetation in their home range. Elephants occupied lower elevations when vegetation activity was high, whereas they retreated to the evergreen forest at higher elevations while vegetation senesced. Elephant home ranges decreased in size, and overlapped less with increasing elevation.

Conclusions: A recent hypothesis that ungulate migrations in savannas result from countervailing seasonally driven rainfall and fertility gradients is demonstrated, and extended to shorter-distance migrations. In other words, the trade-off between the poor forage quality and accessibility in the forest with its year-round water sources on the one hand and the higher quality forage in the low-elevation scrubland with its seasonal availability of water on the other hand, drives the relatively short migrations (the two main corridors are 20 and 90 km) of the elephants. In addition, increased intra-specific competition appears to influence the animals' habitat use during the dry season indicating that the human encroachment on the forest is affecting the elephant population.

Keywords: Human-elephant conflict; NDVI; Rainfall; Remote sensing; Seasonality.

Figure 1

Daily rainfall at the Marsabit meteorological station (vertical bars), observed MODIS-NDVI at 16 day intervals (solid line), and NDVI estimated from the rainfall observations using a random forest model (dashed line).

Figure 2

Descriptive statistics of the annual growing periods in the Marsabit area as a function of elevation. a) The number of growth periods per year, estimated as NDVI > 0.45, and b) the mean length of these seasons, and associated standard deviations.

Figure 3

NDVI and elevational (y-axis, meters) migration of the ten elephants equipped with GPS-collars in the Marsabit area. Color map in the background shows NDVI as a function of height and time (x-axis, in days since the start of the experiment). Dots show the observed elevations of the elephants and the bold lines show the predictions of 3 models. Red line marks the pure surf model (Table 1 model 1), black line marks the individual surf model (Table 1 model 2) and cyan line marks the most justified model we found, which include linear modification of a single "Surf"-NDVI with linear effects of positive and negative temporal changes to NDVI (Table 1, Model 8). Each modeled individual is shown is a different panel. The bottom right panel shows the overall goodness of fit of two of the models, model 8 in cyan and model 2 in black.

Figure 4

The Marsabit study area and recorded elephant movements. (a) with agricultural areas (shaded and outlined in red), major roads (bold lines), minor roads (thin lines), and elephant locations recorded in (b) December 2005–2008 on a background of NDVI measured between December 19 and 31 December 2006, and (c) July 2006–2008 on a background of NDVI measured between 12 and 28 July 2006, draped over a digital elevation model.