Ed and Jim,

I took a stab at incorporating the USFWS dozer trawl data from the Illinois River that Ed sent me on 8/17/23 into my existing workflow for analyzing Silver Carp relative abundance in relation to Silver Carp body condition (Wr). For now, I have not explored the relationship of Ed’s data to native fish Wr, but that could be added if we want to expand the full breadth of my Multi-agency Monitoring (MAM) analysis to this dozer trawl data.

Silver Carp length distributions by gear type

One potential difference between the MAM and USFWS datasets that may influence length-derived relationships is the size-selectivity of the two gear types, electrofishing and electrified dozer trawl. Below, I compare the two size frequency distributions. Generally speaking, both gears capture the same range and relative frequency of total lengths above the total length threshold (200 mm) that I had been using to identify likely adult Silver Carp in my previous analyses. The dozer trawl seems to be relatively undersampling the large ‘likely juvenile’ peak below 200 mm total length that is present in the electrofishing data.

Regardless of the performance below the ‘likely adult’ threshold, there doesn’t seem to be a large discrepency in distributions of total length between the two gears. Patterns among pools also appear similar between gear types.

Fig. 1. Length frequency distributions of Silver Carp captured by (A) MAM electrofishing and (B) USFWS dozer trawl.

Fig. 1. Length frequency distributions of Silver Carp captured by (A) MAM electrofishing and (B) USFWS dozer trawl.

Silver Carp length distributions by major river strata and gear type

After filtering to just ‘likely adults’ (>200 mm) we should consider comparing mean sizes between gear types within each pool. Upriver, these means are similar between gear, but downriver electrofishing catches smaller fish, especially in the backwaters of La Grange. Variability around those means (lines are 95% confidence intervals) are also generally similar, though small differences between the gears occur by pool. For example, in the Alton pool, dozer trawl estimates of mean total length have larger variability than e-fishing. But in Marseilles pool the opposite is true.

Overall, I don’t think these minor differences in annual, pool-wide estimates of total length will impede comparing the two datasets.

There do not appear to be consistent differences of mean total length between major river strata, indicating the relative sampling effort in each strata type shouldn’t be an issue for modeling.

Fig. 2. Strata-specific estimates of mean total length (mm; with 95% confidence intervals) of Silver Carp captured by (A) MAM electrofishing and (B) USFWS dozer trawl.

Fig. 2. Strata-specific estimates of mean total length (mm; with 95% confidence intervals) of Silver Carp captured by (A) MAM electrofishing and (B) USFWS dozer trawl.

Silver carp length:weight relationships by major river strata and gear type

Another important length-derived metric worth comparing is the relationship between length and weight. These relationships will be the basis for our body condition (Wr) calculations. Large differences in the form of these length:weight relationships by major river strata could indicate we should apply strata-weighting to the Wr calculations in a way similar to our strata-weighted estimates of relative abundance. However, there doesn’t appear to be any such large differences, at least not where one strata is consistently different among pools. For now, I won’t use any strata-weighting for the Wr calculations.

Between gear types, the relationships also look very similar.

Relationships were modeled using generalized linear modeling of the form Total Length ~ Major stratum * Pool + Year, using a Gamma family distribution and log link function.

Fig. 3. Strata-specific relationships between Silver Carp total length (mm) and weight (kg) captured by (A) MAM electrofishing and (B) USFWS dozer trawl. Abundance and biomass estimates are weighted by relative proportions of available area sampled within each major river stratum.

Fig. 3. Strata-specific relationships between Silver Carp total length (mm) and weight (kg) captured by (A) MAM electrofishing and (B) USFWS dozer trawl. Abundance and biomass estimates are weighted by relative proportions of available area sampled within each major river stratum.

Comparing Silver Carp relative abundance (CPUE) or biomass (kgPUE) to Silver Carp body condition (Wr)

Finally, we can compare Silver Carp relative abundance (CPUE) or Silver Carp biomass (kgPUE) to Silver Carp body condition, using design-based, annual, pool-wide estimates. I plotted predictions from a generalized linear model of the form Wr ~ Relative Abundance (or biomass) + (1 | Pool). Adding Year as a fixed effect did not result in a significant estimate nor did it improve AIC/BIC.

It may not be useful to directly compare the x-axis values of each plot to the other, as the two gears are probably not comparable. For CPUE, I standardized the MAM e-fishing data to a 15-minute shocking run and the USFWS dozer trawl data to a 300 second trawl. Interestingly, the range of the values appears to be about the same between gear types (~1.5 units of effort), which does inspire confidence that we can compare the trends in a similar way between the two gear types.

It seems a stronger relationship is present within the MAM electrofishing data as compared to the USFWS dozer trawl data. Two outlier groups exist within the MAM electrofishing data: all years of the Starved Rock pool, and 2022 in the Marseilles pool. Removing either (or both) of these outlier groupings lowers the p-value of the generalized linear mixed-effect model (GLMM) enough to achieve statistical significance at the alpha = 0.05 level.

Starved Rock pool estimates also appear grouped in the USFWS dozer trawl data, though their removal does not improve the relationship to the point of a significant GLMM estimate.

In general, correlation is stronger between Silver Carp relative abundance and Wr than it is between Silver Carp biomass and Wr.

Fig. 4. Relationships between Silver Carp relative abundance (CPUE; A, B) or biomass (kgPUE; C, D) and Silver Carp body condition (Wr). Displayed p-values are from generalized linear mixed-effects models (GLMM) using a Gamma family distribution and log link function.

Fig. 4. Relationships between Silver Carp relative abundance (CPUE; A, B) or biomass (kgPUE; C, D) and Silver Carp body condition (Wr). Displayed p-values are from generalized linear mixed-effects models (GLMM) using a Gamma family distribution and log link function.

Conclusions

Overall, I’m not seeing the same strength of trend in the USFWS dozer trawl data that I am pulling out of the MAM e-fishing data. Some patterns are similar, such as Starved Rock being an outlier grouping with particularly high Wr despite a high relative abundance. The general form of a negative relationship is there in the dozer trawl data, but there’s too much noise for it to be statistically significant. What little relationship exists is HIGHLY driven by the Marseilles pool. Without Marseilles, there really is no relationship at all. For the MAM e-fishing data, the data is much more mixed and a trend exists more-or-less throughout the range of relative abundance values.

This method of using design-based (i.e. strata-weighted), annual, pool-wide estimates of relative abundance and Wr for ‘likely adult’ (i.e. > 200 mm) individuals is by no means the ONLY way to analyze this data. My use of GLMMs allows me to control for the random effect of pool, but I could be convinced to drop that or move it into the fixed effects.

I also experimented with the two datasets by combining the e-fishing and dozer-trawl estimates, but no grand revelations were reached. I basically added the e-fishing CPUE to the dozer trawl CPUE for a sort of combined-gear estimate, and then averaged the two Wr values for each annual, pool-wide estimate. It would have been a pretty unorthodox approach, but it didn’t seem to add any value so I didn’t present it here.

I’d be interested to see how you’ve approached the question, Ed. Let me know if you have any questions or would like to see this presented differently.