The spatiotemporal response of mammal communities to pulse-driven resource availability in the Kalahari data
The objective of this study was to detect pulse events at a forage patch scale and to identify the subsequent responses from various mammal functional groups of mammals in the arid Kalahari Desert of South Africa. The aim was to, therefore, understand how the Kalahari functions on a finer scale, particularly with increased rainfall variability in the drier years. We predicted the green-up (pulse event) to be synchronized over large areas with extreme rainfall, while we expected low synchronicity and very patch green-up in dry years. As ephemeral plants respond to rain showers within 2-3 weeks, we predicted the following responses from the animal functional groups would be detectable within ~8 weeks after the green-up: 1. There would be an increase in abundance of large antelopes. 2. There would not be an increase in abundance of small antelopes after the pulse event. 3. There would not be an increase in abundance of carnivores after the pulse event. 4. There would be an increase in the meso-carnivore abundance. 5. There would be an increase in rodent and lagomorph abundance. 6. As our cameras are in the dunes, we predict that during a period of normal rainfall, there would be a decrease in observations as animals may move to more attractive areas. We gathered the data from 42 cameras located in Tswalu Kalahari Reserve (n=18), Kgalagadi Transfrontier Park (n=12) and Khomani San Community Land (n=12). We used greenness index values obtained from the phenopix package in R, and subsequently used relative abundance index to compare functional group abundance 4 weeks before the pulse event and 8 weeks after the pulse event, to determine if there was a response. We used two survey seasons, October to May in 2018-2019 and 2019-2020. In this study we developed a novel way to detect pulse events in arid systems using bycatch from camera trap data. We found clear patterns of green-up and a clear difference in the green-up/resource availability between season 18-19 and 19-20. We also found some differences in the mammals RAI between season 18-19 and 19-20. We were able to link the pulse events to the short-term response of large antelopes, in Kgalagadi and the Khomani San. We did, however, not see any other responses to the green-up, which with the limitations of our data suggest that the responses are likely to be longer term or occurring at a larger spatial scale. This has allowed for a deeper understanding of animal responses in unpredictable arid environments. Climate change threatens systems such as the Kalahari with rising air temperatures and further precipitation unpredictability. Droughts are to increase in magnitude and frequency. With this increase in system variability, it is crucial that arid landscapes are managed effectively, both short-term and long-term. To be able to manage the system, it is of utmost importance to further our knowledge of the Kalahari system. This study has provided a starting point to do so.
Steps to reproduce
One photo per week, from each camera, from the beginning of October to the end of May, was analyzed in phenopix to obtain greenness indices for all 42 cameras. To test for a difference in the abundance of wildlife before and after a pulse event, we needed to specify a specific point in time when a pulse occurred. In order to do this, we generated a three-week moving average of the GI values for each camera trap site to determine averaged trends in GI per camera. We then identified the onset of a pulse event as the first week during which the three-week moving average was above the seasonal mean GI. We subsequently tallied the total number of individual observations for each of the functional groups in the four weeks before and the eight weeks after the onset of the pulse. The four weeks leading up to the pulse event includes the rainfall event that prompts primary productivity, but other than that shows the base conditions of the Kalahari – limited resources and dry conditions. The eight weeks after primary productivity begins, was chosen based on the predictions on when functional groups would respond to the primary productivity. Most functional groups were analysed by running a GLM with a Poisson family and log link function, with treatment, season, offset and a random intercept per camera as covariates. The model for the AS group in ‡Khomani San failed to converge due to a large number of zeroes resulting in overdispersion. We thus analysed this data using detection/non-detection of the functional group at each camera trap in a GLM with a binomial family and logit link function.
National Research Foundation
National Research Foundation
Foundational Biodiversity Information Programme