Food web properties and the type of invasive species make the ecosystem vulnerable to invasion - web files and analysis code
Data files
Aug 04, 2026 version files 8.39 GB
-
Data_randomweb_withinvasive_json.zip
51.09 MB
-
Data_randomweb_withoutinvasive_json.zip
53.53 MB
-
README.md
6.29 KB
-
Results_randomweb_withinvasive_json.zip
4.53 GB
-
Results_randomweb_withoutinvasive_json.zip
3.76 GB
Abstract
Invasive species have long been acknowledged as potentially severe dangers to native ecosystems. Although some work has also been done using food web approaches on empirical food webs, more information is still needed to shed light on cascading effects of invasive species in aquatic ecosystems. Here, we used a modified niche-network approach to generate food webs with and without invasion, Then, we used allotropic network modelling to simulate 200 years of species biomass changes with and without the invader in the ecosystem. We found that the biomasses of highest trophic levels of species (fish) changed the most and the response was often a decline in the biomass. Among fish, old age groups were most affected. Network depth and the magnitude of ecosystem-level changes caused by invasion predicted how several effect the invader had. Both the properties of native species and the invader, e. g. the tropic level and both direct centrality and indirect centrality, reflecting the cascading effects, affected to the outcome. High topological similarity between invader and the native species was also a predictor of negative impact to the commercially utilized stocks of large fish.
https://doi.org/10.5061/dryad.t76hdr885
Description of the data and file structure
The input webs and simulation results are in zipped folders and in a json format.
Data_randomweb_withoutinvasive_json.zip (3275 files)
Data_randomweb_withinvasive_json.zip (3030 files)
Results files are:
Results_randomweb_withoutinvasive_json.zip (3275 files)
Results_randomweb_withinvasive_json.zip (3030 files)
Differences between the number of files with and without invasion stems from the fact that 25 of the invaded food webs could not be included in the final analyses but we, nonetheless, wanted to provide the full raw data.
Naming conventions of files
Each Data and Results json-file contains one web. The names of the webs contain the number of species S (num_species[S]) used to generate the webs, expanded to 59 guilds after fish age-structuring (or 60 with the invader), the target connectance C (connectance[C]) and a time stamp which identifies each web. The files include also trophic species properties detailed in Sävilammi et al. (2026) Table 1 as well as below. Parameters C and S used to generate the webs were S = 50 and C = 0.15.
Data files
Each Data file contains a random food web with 50 species including 47 trophic guilds of producers and consumers, and 3 species of fish with 4 age classes (trophic guilds) each (in total, 47+12=59 trophic guilds) without (Data_randomweb_withoutinvasive_json folder) or with (Data_randomweb_withinvasive_json folder) an added invasive species (47+12+1=60 trophic guilds). Each json file describes web parameters which are listed below:
TrophLevel describes the average (short-weighted), (Williams & Martinez 2004), shortest path, and prey-averaged trophic level of each trophic guild in the web.
Matrix adjacencyMatrix is a 0/1 matrix containing species and age classes in both rows and columns. Each cell contains 1 if the guild in row eats the guild in column, otherwise 0.
B0 is a matrix with the number of rows and columns corresponding to species and age-groups. It describes the functional response to half-saturation density for each link. The value is non-zero only when the adjacencyMatrix is 1.
Guilds is a list which includes trophic guild -specific parameters: label is an abbreviation for the guild (e.g. P01=producer 1, C01=consumer 1, F1A0=fish 1 age 0, consumernew=the invasive species). Name is a long name for the guild, e.g "Producer#1", "Consumer#1, "Fish#1,0yr". Type describes the metabolic type: "Producer", "Consumer" or "Fish". Binit is the start biomass in weight units. The invasive species has value 1000 units. Diss_rate is not used in this model and thus equals 0. Igr is the intrinsic growth rate of the producer guilds. Mbr is the basal metabolic/mortality rate for consumers and fish. Avgl is the average length for fish, calculated through the growth model. Lw_a and lw_b are allometric rates for length-to-weight ratio of fish. Age is the age group 0-3 for fish, otherwise 0. S is the self-limitation rate of the producers, here set to 0.2. C is the predator interference constant, Beddington–DeAngelis-type. Catchable is 1 for the fish which can be fished, otherwise 0. F_a is the assimilated carbon for growth, here 0.4 (Boit et al. 2012). F_m is the assimilated carbon for basal consumption, here 0.1 (Kath et al. 2018). Species is numbering for the species. Hatchery is 1 for hatchery-origin individuals (here 0). Invest is the breeding investment rate. Centerofrange, range and nichevalue correspond to niche-model parameters (Williams & Martinez 2000). MaxAge on is the maximum age of the fish. Pmat is the maturity ogive for fish.
D matrix is the predator interference for feeding links (1 if no difference between the preys).
Y matrix is the maximalmass specific feeding rate.
Q matrix is the functional response for Hill exponent.
E-matrix is the assimilation efficiency for links. 0.45 for herbivores, 0.85 for carnivores (Brose et al. 2006).
K is the carrying capacity for basal producers. 540000 weight units. Temporal variability is not used in these simulations.
Mvec describes the masses of the guilds.
NGrowthdays is the length of the growth season in days per year (here 90).
T_init is the start time of the simulation (here 0).
Tspan: the time steps during a growth season (here 0-90).
Ltspan is the length of tspan (here 91).
Start_fishing, end_fishing, F and gear are fishing parameters, not used here.
End_recovery is the length of the burn-in time (here 1000 growth seasons).
GuildInfo is an internal indexing list of the trophic classes of each guild.
Varstoflip is a technical helper variable.
Results files
The associated results files contain the data during the simulation steps during the test (invasion) and control period. The associated Results files contain biomass changes across 200 simulated years (columns) with and without an invasive species (variable B). The biomasses are given as densities with the unit being micrograms of carbon per cubic meter of water. Biomasses are relative to primary productivity carrying capacity and, thus, their relative changes is of interest here. Before simulations, the random webs were run for 1000 years of a burn-in period, to ensure they do not collapse but are biologically feasible. Variable GF describes the growth by breeding for the fish.
Variable G describes the biomass allocated to predators by feeding, and L biomass flux by cannibalism. After random food web construction, a script to extract biomass changes of fish and several web- and invasive-species-related network topological properties is associated. Top trophic levels (fish) were most severely affected (tendency for a negative biomass change) in response to invasion. Several of the properties of the web, the trophic species, and the invasive species itself affected to the fish stock decline.
Software files hosted by Zenodo
An R script to extract the results is attached (R v. 4.5.1).
The software files:
Sävilammi_simulation_codes.zip 170.24 KB
Sharing/Access information
We used a modified niche-network approach to use two simple parameters (connectivity and the number of species) to generate artificial food webs with trophic species which have random niche values, prey ranges, and centres of the prey ranges. We also added age groups to the top-three trophic levels for each web to represent fish. With these random parameters, food webs with natural resembling properties were generated. We also added an ivasive species with random niche value, prey range, and center of prey range.