Understanding arid‐region waterbird community dynamics during lake dry‐downs

These data were collected to explore changes in the bird community associated with Lake Ngami, Botswana, through successive drydown periods. Our analysis shows significant shifts, driven partially by changes in water level, in the species composition of the bird community over the period of study. The data set contains standardised half-hour point counts for the bird community of Lake Ngami, Botswana; and R code used for community-level analyses of the resulting time series. Counts were undertaken every 4 months from October 2007-July 2009.

We used the data to test three theoretical predictions: simplification of the bird community over time due to a reduction in habitat area and concurrent niche loss; large fluctuations in densities of mobile, opportunistic species; and high variance in predator and prey abundance. Despite temporal variance in species accumulation, we observed no obvious simplification of the bird community over time. There were distinct but consistent groupings of abundance and composition across transitional stages in lake water levels. The data do show some rapid shifts in functional composition, such as loss of deepwater foragers; winners and losers also occurred within foraging guilds.

Methods
We quantified changes in species diversity and the functional composition of the waterbird community over two iterations of annual drying-down and flooding. Each sampling period involved five days of standardized point counts at between 8 and 13 points. Points located at the water's edge were each visited four times, at different times of day, over each 5-day counting period. The total sample size was 228 point counts. Each count used a ten-minute habituation period followed by a half-hour counting period by stationary observers, during which all birds seen within a 150m radius in front of the observers were recorded. All counts included at least one highly skilled observer and were mostly undertaken by the same group of people. The counts contain four standardized samples for each point for 8-13 different points, and a minimum of 32 point counts (16h observation time) per sampling period.