Ecosystem age drives food web architecture of glacier retreat-formed fishless ponds in Greenland

Here, we studied the topology of food webs in newly created Arctic ponds formed by deglaciation, analyzing potential drivers such as environmental variables in 16 fishless ponds of contrasting ages (8 young, <50 years; 8 old, >150 years) located on the rapidly retreating west coast of Greenland. Using gut content analysis of zooplankton and benthic macroinvertebrates, we constructed food webs and examined how food web metrics related to pond age and environmental variables. We also tested for differences in beta diversity and nestedness of prey and consumers between pond age classes. In young ponds, food items of zooplankton and macroinvertebrates represented a subset of those found in old ponds. Food webs in old ponds exhibited higher taxonomic richness, greater linkage density, and lower connectance, indicating greater stability. Pond age was the main explanatory factor for most food web metrics, followed by productivity proxies (total nitrogen and phytoplankton chlorophyll-a). While food webs in old ponds were not significantly associated with environmental variables, in young ponds, connectance and trophic niche overlap were significantly related to TN and Chl-a. These contrasting patterns suggest that food web topology becomes more complex and less dependent on environmental conditions as ponds age.

Methods
Samples were obtained during a single survey of ponds located on the western coast of Greenland. Pond age was estimated at the landscape level using current and historical ice-sheet maps and aerial photographs measuring distance to the ice front, rather than at the level of individual systems. Half of the ponds (n = 8) were located in an area where the glacier had retreated within the past 50 years (“young ponds”), while the other half were situated in an area deglaciated for at least 150 years (“old ponds”).

Stomach contents of macroinvertebrates and zooplankton were analyzed to reconstruct food webs. Using the relative abundance of food items by taxon and by system, we built diet matrices for each pond (file: data1.csv) to calculate trophic network metrics (connectance, linkage density, and trophic niche overlap). For graphical representation of networks, we also constructed edge lists for each pond, where each row represents a trophic interaction from a consumer to a resource (file: data2, folder with spreadsheets for network reconstruction).

Based on consumer identifications by pond, we created a presence/absence matrix of consumers across sites (file: community cons.csv). Similarly, based on the food items found in guts, we created a presence/absence matrix of prey items across sites (file: community2.csv). From these matrices, we calculated overall beta diversity (βsor) and its components (turnover, βsim, and nestedness, βnes).

Morphometric and environmental variables were measured in each pond: surface area and maximum depth (measured in situ), water pH, temperature, and electrical conductivity at the central point, as well as water samples for chemical analyses. Indicators of trophic state included total phosphorus (TP, μg L⁻¹), total nitrogen (TN, μg L⁻¹), and phytoplankton chlorophyll-a (Chl-a, μg L⁻¹) concentrations, as proxies of potential productivity. Differences between pond age groups were tested using two-tailed Mann–Whitney–Wilcoxon tests at α=0.05 (file: Environmental variables.csv).

Relationships between food web metrics (number of food items, maximum trophic level, connectance, linkage density, and trophic niche overlap) and environmental, age, and morphometric variables were analyzed using Generalized Linear Models (GLM) (file: GLM_analysis.csv).

All analyses were conducted in R using standard packages and functions, as described in the methodology section of the article. The accompanying script provides an example of each type of analysis; full details are available in the published paper.