Computational protocol: The underestimated role of temperature–oxygen relationship in large‐scale studies on size‐to‐temperature response

Similar protocols

Protocol publication

[…] The originally studied area consists of fourteen streams in the Hengill region of Iceland, the water of which is differently altered by geothermal warming and ranges from 5 to 25°C (Adams et al., ). These streams were already the subject of research on different aspects of the effects of temperature: community structure and trophic interactions (Woodward et al., ), primary production (Gudmundsdottir et al., ), and individual vs. ecosystem‐level response (O'Gorman et al., ). The temperature was measured in the streams in August 2008, along with diatom sampling (Adams et al., ) and it was recorded on the same day in all streams (Woodward et al., ). To determine if the streams grouped according to specific conditions, we conducted a PCA analysis with a correlation matrix (CANOCO 5; Ter Braak & Šmilauer, ) using all the physico‐chemical parameters collected in the supplementary materials (table ) of Adams et al. (). In a second step, we removed temperature and oxygen (the factors of our main interest) from the analysis to check if the other parameters differentially affected the stream aggregation. We also checked the correlation matrix (Pearson's correlation coefficient) of all the abiotic parameters (STATISTICA 10, Statsoft, ). [...] For this analysis, we used the species database from the Supplementary Materials in Adams et al. (; table S10), containing 37 diatom species analyzed for body size in an original article. The product of length and width (original values provided by the authors) was used as an approximate size measure (in the original work, the authors also estimated diatom area, but they used a genus‐specific standard geometric shape in their calculations). We checked the diatom temperature‐size relationship of each stream group, arbitrarily separated according to the PCA plot, with a linear regression analysis (data analysis using one statistical model with stream group and species as model factors was not possible because of the uneven distribution of species; only six species from an original species database were present in all three stream groups, and only one of them was present in all streams). Additionally, for each species within a stream group we estimated a coefficient of regression for the size‐temperature relationship and we conducted the effect‐size analyzes of grand mean for r, using a meta‐analysis tool (OpenMEE software, Wallace et al., ), for each stream group separately. We used a fixed‐effect model for Fisher's Z‐transformed effect size. […]

Pipeline specifications

Software tools Statistica, OpenMEE
Applications Miscellaneous, Phylogenetics