Appendix D. Description of the data sets on species richness and data analysis.
Datasets reporting oxygen clines and patterns in species richness
To test the efficacy of the OSI to explain patterns in species richness, we reviewed the published literature on altitudinal clines in species richness, where both oxygen and temperature were measured and where potentially confounding effects of covariates (e.g. stream width, substrate composition or human impact) with altitude was minimal. Only two datasets for mountain streams were found that meet these criteria; one in southwest Colorado, spanning 2,315–3,190 meters above sea level (m a.s.l.) (Perry and Schaeffer 1987) and the other in the Ecuadorean Andes, between 2,600–3,980 m a.s.l. (Jacobsen 2008). In both datasets, species richness decreased with altitude and increased with increasing temperature. To test the generality of the proposed index of oxygen supply we also reviewed the literature for additional datasets where species richness decreased with increasing temperatures. Two other datasets were included, both featuring a comparison between organically polluted and unpolluted sites. The first of these additional datasets was obtained from the Bolivian Altiplano (3,800–4,000 m a.s.l.) where oxygen supply is already limiting and thus the effects of oxygen are expected to be more important relative to other environmental conditions (Jacobsen and Marín 2008) (see also below). The second data set featured a comparison between polluted sites and upstream, reference sites in Patagonian streams, Argentina (Miserendino et al. 2008). In these additional datasets oxygen is (periodically) lowered due to organic pollution, and poor oxygen conditions coincided with higher temperatures. Species richness decreased with increasing temperatures, in contrast to both datasets with an altitudinal cline. Our oxygen supply index consistently explained richness patterns in all four datasets, independently of whether species richness is positively or negatively related to temperature.
Analyses of the relationships with species richness
The ability of altitude, temperature and several measures of oxygen, including the OSI and rOSI, to explain patterns in species diversity was tested using linear regressions. Normality of data was confirmed using a Kolmogorov Smirnov test. Subsequently, pair-wise likelihood ratio tests (LRT) were used to determine if one variable was a significantly better predictor over another, using the formulae given in Johnson and Omland (2004) to calculate statistics from standard regression output. The explained variation in richness or body size was assessed using Pearson’s R. All statistical analyses were performed with SPSS (17.0, SPSS Inc., Chicago, Illinois, USA). Since rarefied richness was not reported in the original papers, rarefied richness was calculated for the Colorado mountain stream dataset (Perry and Schaeffer 1987) and the Patagonian streams (Miserendino et al. 2008) using Biodiversity Professional Beta 1 (N. McAleece, 1997). Site 1 was an outlier in the dataset on Colorado mountain streams. This site was at the mouth of the creek and environmental conditions deviated strongly from those found at the other sites both with respect to substrate (much smaller) and sediment loads (much higher) (Jim Perry, pers. comm.). This will have had strong influences on the species assemblage, as increased sediment loads will have contributed to oxygen depletion and indeed the species richness was very low at this site. However, since oxygen consumption in the sediment is not reflected in the BOD5 measurement, the degree of oxygen depletion and its relationship to the low species richness could not be quantified. Therefore this site was not included in further analyses.
Since the lower extreme (poor oxygen conditions) is assumed to be the most critical, both OSI and rOSI were calculated for those temperatures where they reached the lowest values (i.e. OSI was calculated for minimum temperatures and rOSI for maximum temperatures). When actual minimum or maximum temperatures were not reported we used the mean plus or minus the square root of the standard deviation (to avoid generating negative values). To derive minimum oxygen concentrations from the reported BOD5-values for the Colorado mountain stream dataset, we assumed that oxygen consumption followed a hyperbolic curve and took half (1/2) the BOD5-value as an estimate for oxygen consumption in a single night. Different values (1/3, 1/5) did not qualitatively alter the results.
Jacobsen D. (2008) Low oxygen pressure as a driving factor for the altitudinal decline in taxon richness of stream macroinvertebrates. Oecologia 154:795–807.
Jacobsen D. and Marín R. (2008) Bolivian Altiplano streams with low richness of macroinvertebrates and large diel fluctuations in temperature and dissolved oxygen. Aquat. Ecol. 42:643–656.
Johnson J. B. and Omland K. S. (2004) Model selection in ecology and evolution. Trends in Ecol. Evol. 19:101–108.
Miserendino M. L., Brand C., and Prinzio C. Y. D. (2008) Assessing urban impacts on water quality, benthic communities and fish in streams of the Andes mountains, Patagonia (Argentina). Water Air Soil Pollut. 194:91–110.
Perry J. A. and Schaeffer D. J. (1987) The longitudinal distribution of riverine benthos: A river dis-continuum? Hydrobiologia 148:257–268.