DOES SCALE AFFECT ECOLOGICAL MODEL PREDICTIONS? A TEST WITH LAKE RESPONSES TO FERTILIZATION

Ecosystem ecologists often face the challenge of predicting long‐term consequences of perturbations such as nutrient enrichment from shorter term experiments. In such experiments, the spatial and temporal scale at which data are analyzed can directly impact extrapolations to larger scales. Here, we assess the level of data resolution required to answer qualitative and quantitative questions about primary production in a set of lake ecosystems. We tested the ability of 40 models containing variable levels of spatial and temporal complexity to (1) fit the relationship between light and primary productivity in a set of 14 British Columbia (Canada) lakes that were part of a large‐scale fertilization experiment, and (2) predict annual primary production. The experimental data had previously been averaged across fertilization treatments for analysis, effectively ignoring spatial and temporal variation. In the limnological literature, data from whole‐lake experiments are often analyzed for each lake to account for among‐lake differences, or analyzed for each lake in each year to account for both lake effect and interannual variation. Using an information‐theoretic approach, we tested these three models (“fertilization status only,” “lake only,” and “lake and year”) against models that included less and more spatial and temporal partitioning of the data. We fit a Monod function to light and primary productivity at 40 spatial and temporal levels of data resolution and ranked the model fits using the second‐order Akaike Information Criterion (AICc). The “fertilization status only” model ranked 37th out of 40 models tested, the “lake only” model ranked 27th, and the “lake and year” model ranked 25th. The top‐ranked model partitioned the data by lake, year, fertilization status, season, and sampling station, and fit the data substantially better than other models. We then calculated primary production with independent light data and the fitted model parameters to compare the top three models from our AICc ranking and the “lake only,” “lake and year,” and “fertilization status only” models. The predicted production differed depending on the model, but all models predicted higher net primary productivity in fertilized lakes. Model selection is therefore important for quantitative predictions, but not necessarily for qualitative assessment of nutrient limitation.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.