FEATURE ARTICLE
Water, Migration and the Serengeti Ecosystem
Understanding the mechanisms that control the timing of wildlife migrations may prove vital to successful management
Eric Wolanski, Emmanuel Gereta, Markus Borner, Simon Mduma
Water Quality
We measured water pH, salinity, dissolved oxygen and temperature (20 centimeters below the surface) at 49 stations in most rivers and other water bodies in the park every three months from April 1996 to April 1999. We also measured the vertical distribution of these parameters at some of these sites, estimated visibility and collected undisturbed samples for microscopic analysis of suspended matter.
Although water quality may be reasonably good in rivers during the wet season, the picture changes markedly during the dry season. Stagnant ponds are heavily used by wildlife, and the water reflects this abuse. Animal dung generates algal blooms in the surface waters of most ponds, and wildlife tramples the bottoms, further increasing turbidity. As a result, visibility is commonly less than 3 centimeters, and about 40 percent of ponds have visibility of less than 1 centimeter. Except for the top few centimeters of water, there is complete darkness.
Suspended matter consists of macro-aggregates made of organic matter (500-5,000 micrometers in diameter) and small mud flocs (less than 100 micrometers in diameter). Because of the turbidity, the water is highly thermally stratified (2 degrees Celsius per meter), which inhibits vertical mixing and aeration of deeper water. Aeration of bottom waters takes place only when hippos stir the water (see Figure 7) or when warm layers develop along the bottom (perhaps as a result of decaying animal dung and other organic matter). Thus dissolved oxygen levels also have a strong vertical gradient, varying diurnally from about 1.1 to 8 parts per million (ppm) at the surface and 0 to 4.5 ppm at the bottom. Bottom waters in our study were anaerobic about 10 percent of the time.
Among the data we collected on water quality, salinity proved to be perhaps the most interesting. In general water salinity is highest in the grasslands in the southern portions of the park. Most water was very saline in the southern grasslands during the 1996 wet season, whereas sweet water prevailed at the same time in the north. In April of 1996, at the peak of the wet season, salinity was commonly 5–15 parts per thousand in the southern grassland, 0.5–20 in the Seronera River (a small tributary to the Grumeti River), 1.8–5.8 in the Mbalageti River, 0.5 in the Grumeti and less than 0.1 in the Mara River (see Figure 5).
During the dry season, evaporation dries out the water holes of the southern grassland, leaving salt behind. The 1997 wet season was drier than that of 1996, resulting in less dilution and even higher salt concentrations. The alkaline lakes Ndutu and Magadi reached a wet-season minimum salinity of more than 20 in 1996 and 1997, but, with the increased rains of 1998, these levels dropped by half. Note also that the salinity in the Seronera River during the 1998 wet season was relatively constant along its length, suggesting that flooding had homogenized the water along the river's course.
Although salinity can be quite high in the headwaters of the Seronera and Mbalageti rivers, salinity decreases rapidly downstream in both rivers to a level of less than 1 in their lower reaches. The salinity shows much variability during 1996 and 1997 along the course of the Seronera River (the most heavily sampled body of water in our study) at 18 stations. When salinity exceeds 4, no grass grows along the banks of the rivers, and, when it drops below 2, the vegetation changes from grassland upstream to wooded savanna downstream. (Coincidentally, rice cannot be irrigated with water whose salinity is greater than 2.) We propose that salinity variations on decadal time scales may control the discontinuity between grassland and wooded savanna.
Furthermore, we argue that excessive salinity may be the trigger that starts the migration of wildebeests and zebras from their wet-season range in the southern grassland to the northwest. During 1996 and 1997, we observed the animals beginning their migration when there was still plenty of forage and surface water in the area to sustain the herds. The salinity of the water, however, was very high. No data are available on the length of time that wildebeests and zebras can survive on such high-salinity water. Sheep, however, have an upper limit of about 10 over the course of a few months. In the arid Kalahari-Gemsbok National Park of South Africa, groundwater is pumped to ponds for wildlife. Where the water is hard and saline, the wildebeests leave; where it is sweet, they stay.
Based on this hypothesis, we have developed a model that predicts the timing of migration as a function of salinity. We use the salinity levels in Lake Magadi to provide a single number that is roughly proportional to the numerous water sources in the southern grasslands. Although the Seronera and Mbalageti rivers are more important water sources and are considerably less saline overall than Lake Magadi, their salinities appear to vary with good agreement. The model begins on April 15 of each year and calculates a threshold value for salinity that will coincide with the day the migration passes by Seronera (halfway between sites 1 and 18 on the Seronera River).
The model takes into consideration initial salinity in Lake Magadi, water depth (which begins at the same depth each year because of a sill), the evaporation rate and rainfall. Observations in 1996 indicated that 30 is the threshold value for migration from the southern grassland to the woodlands. Thus we were able to predict the migration timing for 1997 through 1999 by estimating when the waters of Lake Magadi would reach that value. In each case, our prediction (made early in April) for the migration was within one week of being correct (late April in 1996, early May in 1997, late July in 1998 and late June in 1999, when our estimate was only one day off).
It is quite likely that actual measurements of salinity in a sample body of water will prove unnecessary in the long run to predict migration. Because rainfall and salinity are inversely proportional, we think that a rainfall-runoff model could predict salinity and thus migration. Moreover, a previously developed rainfall-runoff model for the Mbalageti River could be used to estimate previous salinities, and hence migrations, over the past 38 years. These data could be related to historical information from aerial observations of migration routes and of vegetative patterns to construct a deterministic model of the Serengeti ecosystem.
» Post Comment