We often think of a desert as being an exposed ecosystem with a highly variable, severe environment where variation in precipitation is critical. In contrast, we view a lake as a more constant environment well buffered from thermal change by the mass and heat capacity of water and from biological invasions owing to isolation. How can such qualitative presumptions be tested quantitatively?
Comparison of different ecosystems in a non-trivial fashion is often difficult because there are few common entities. Deserts and lakes provide a classic example of this. It is difficult to compare fish and small mammals, zooplankton and ants, forbs, and phytoplankton. The more imaginative can think of analogous metrics in structure of food web assemblages and processes such as energy flow and biogeochemical cycling. Yet, we do not have a means of quantitatively comparing the temporal and spatial variability of desert and lake ecosystems.
An intersite activity is in full swing to explore new metrics for comparing similarities and differences in the temporal and spatial variability of diverse ecosystems. Variability is common to all ecological systems and is a feature which is thought to be important by ecologists. Although ecosystem variability is not well understood. we expect that a serious analysis of it can provide insight into forces affecting ecosystem processes as well as a meaningful measure with which to compare different ecological systems. Here we provide some preliminary results comparing the variability of two ecosystem types: a southwestern U.S. desert (the Jornada Desert LTER Site in New Mexico) and northern temperate lakes (the North Temperate Lakes LTER Site in Wisconsin).
To formalize the analysis, we computed variability estimates from a standard two-way analysis of variance where location in the ecosystem (different geomorphic units at Jornada; different lakes at Wisconsin) and year of observation are the two treatment effects. We included a wide variety of climatic, edaphic, and biological parameters in our analyses (56 parameters from Jornada, and 102 from North Temperate Lakes). Three estimates of magnitude of variability were obtained for each
parameter: variation among locations, variation among years, and variation owing to the combination of measurement error and interaction between year and location which we called variation owing to “other’. Data were relativized by dividing each value in a location by year matrix by the grand mean of the matrix. This transformation was done to dampen the effects of measuring different parameters in different units, We also calculated the proportion of variance explained by location, year, and other from a model II ANOVA.
As most realists might have expected, the answer to whether deserts are more variable than lakes is “it depends on the measure of variability and on whether we consider climatic, edaphic, plant, or animal data” (see figure). However, there are some generalities that emerge in our first statistical analyses using non- parametric paired comparisons.
Among years, deserts are more variable than lakes (see figure). Relative variances among years for deserts are greater than for lakes for climatic and plant data. Also, the proportion (r2) of the total variance due to years is greater for deserts than for lakes for climatic, plant, and animal data. Apparently deserts are less buffered from year to year variability in weather than lakes and this is evident not only in the climatic measures but also in the plant and animal data.
Among locations, lakes tend to be more variable than deserts (see figure). The proportion (r2) of the total variance is greater for lakes than for deserts in climatic and edaphic characteristics, but greater for deserts for animal characteristics. Apparently the greater physical isolation among lakes and the influence of position in the groundwater flow system are more influential than the position along the catena in the desert, at least for abiotic variables.
Among “other”, deserts tend to be more variable than lakes for abiotic variables while lakes tend to be more variable for animals (see figure). The striking feature is that variation owing to other (interaction and error) is much greater for animals in lakes than deserts. We think that the interaction term dominates for animals in lakes because most of the fish in our samples are young of the year and the recruitment of young fishes is notoriously variable among years, but different lakes respond differently in a given year. For example in one lake one species of fish (yellow perch) is dominant while in another lake a different species (bluegill) is -- conditions favoring one species may not favor the second species. Also, the smaller lakes are better buffered from damaging effects of storms on fish eggs and larvae compared to the large lakes. Thus the interaction between year and lake effects could well be real. The greater importance of interaction in the abiotic characteristic of deserts over lakes is less clear. Perhaps differences in water retention characteristics among the desert locations cause the different geomorphic units to respond differently to rain events.
The above analysis and preliminary interpretation appear to provide a means of making meaningful comparison between divergent ecosystems such as lakes and desert. They also cause one to search the back of their mind for processes which might generate the difference in observed patterns of yearly and spatial variation. Some of these confirm our expectations, others we believe will catalyze new research.