This is a summarised version of the Invasive Carp abundance vs. body condition analysis of 2019-2021 MAM data for the January 2023 MRWG meeting. For full, detailed analysis, see the ‘MAM_mission_support_121622.Rmd’ report at https://rpubs.com/mspear2/986104.

The analysis is limited to adult (>200mm) Silver carp and Grass carp, which make up the vast majority of non-native carps in the Illinois Waterway. Because reliable abundance estimates are critical for generating regressions, we only include year x pool combinations where Invasive carps were detected by our sampling.

SPECIES_CODE n percent
BHCP 38 0.05
GSCP 20416 24.56
SVCP 62671 75.39

Invasive carp abundance, both in number and biomass, correlate strongly with Invasive carp body condition. In other words, the fewer carp we have, the fatter they are, which seems to be a classic case of intraspecific competition. This relationship is visually compelling, and in a Genearlized Mixed Effects Modeling approach is statistically significant, even despite an errant grouping of high-abundance/high-condition points, which all come from the Starved Rock pool.

These outliers are even more clear when we plot predictions from our fitted model alongside our raw data. Though the prediction intervals surrounding our model’s predictions are wide, it’s clear that if we found carp that were 15% fatter than the previous year, we could pretty reliably say there were significantly fewer carp, particularly when omitting Starved Rock. This is a pretty impressive relationship, in our opinion, given the limited sample size of three years. Naturally, more years of data will boost the sample size and likely shrink these prediction intervals.

Predicting abundance from condition

There seems to be something particularly interesting about what’s happening in Starved Rock, where fish are in excellent condition despite high abundance estimates. However, we happen to have particularly low confidence that our Starved Rock abundance estimates approximate the true abundance, at least compared to other pools, as evidenced by the Standard Errors of the Means being nearly double the next highest pool…

Even so, just three years of MAM data clearly show that the body conditions of native fishes also correlate negatively with Invasive carp abundance. Similar to Invasive carp body condition in Starved Rock, the native fishes body condition in Starved Rock are outliers of this trend. Because this regression analysis treats the x-value, in this case Invasive carp abundance, as a ‘known’ quantity without error, and because we have substantially less confidence that our abundance estimates in Starved Rock are as close to the true abundance as our estimates in other pools, we’ve eliminated Starved Rock from the analysis of the native fishes conditions. There is definitely something interesting going on in Starved Rock, where again we have fatter fish than we’d expect for how abundant the carp are, and it’s worth exploring further.

That said, we again see a remarkably strong relationship between Invasive carp abundance and the body condition of some native fishes with which we might expect to have strong food web interactions. We have statistically significant, negative relationships for 3 out of 5 species of interest: Gizzard shad, Emerald shiner, and Common carp. We actually have a strong positive relationship with River carp sucker condition, which may be relying more on benthic macroinvertebrates than these other fishes, which are largely planktivorous. Perhaps this is evidence of benthic enrichment from pseudofaeces resulting in mixed directional interactions with different guilds of native fishes?

Plotted data are annual, design-based, pool-wide estimates of Invasive carp abundance and annual means of body condition based on LTRM life history database length-weight regression parameters for each species. P-values represent results of Generalized Linear Mixed Models with pool as a random effect, or where GLMMs did not converge, of Generalized Linear Models with no pool effect.

These limited data already provide evidence for strong interspecific interactions, both antagonistic and synergistic, between invasive carps and native fishes. In theory, well-fit models could generate useful predictions of Invasive carp abundance from native fish body condition, or vice versa. Similar to the relationship between Invasive carp abudnance and Invasive carp body condition, confidence intervals around those predictions are quite wide given our current models trained on only three years of data, but those prediction intervals will likely narrow as sample sizes increase, making prediction more useful. Should invasive carp become increasingly dectable in upper pools, we should gain reliable data points at an increasing rate each year.