The Marais poitevin is a ca. 1,020-km² eutrophic agricultural marshland located on the French Atlantic coast (46°30'N–46°15'N, 1°30'W–0°35'W, Fig. 1). The marshland was reclaimed during the Middle Ages and an 8,200-km long network of lentic drainage ditches has progressively been set up (EPMP 2015). The Marais poitevin is divided into ca. 250 distinct embanked hydrological blocks ranging from 0.05 to 60 km², in which the water regime is controlled relatively independently from the surrounding blocks and managed mainly for agricultural purposes (mainly thanks to hydraulic structures; EPMP 2015). The land surrounding the ditch network is dominated by grasslands, croplands and woodlands (45, 32, and 13% of the hydrological blocks under study, respectively, see below).

Figure 1. 

Localisation of the 11 selected hydrologically independent blocks within the Marais poitevin (delimited by a red line) where crayfish were sampled in specific ditches (blue dots) within dense networks of ditches.

The red swamp crayfish was recorded for the first time in the Marais poitevin in 2003 (EPMP 2015), but it was probably introduced before: it was detected at least 10 years earlier in the adjacent Charente department (Changeux 2003). Since this date, crayfish have colonised the whole marshland. Therefore, they had been present for ca. 10–20 years at the beginning of the study. To investigate whether the water regime is an important driver of the population dynamics of crayfish, we selected eleven hydrological blocks representative of water regime diversity in the Marais poitevin (Fig. 1). Water regime is controlled independently from one block to another because of different water regulations fixed by local management committees. This leads to different hydrological patterns between blocks [see also Crabot et al. (2023)]. The climate is a warm Atlantic one, with mild temperatures and heavy precipitation in winter, and hot, dry summers.

Crayfish were collected in one medium-sized ditch (4–10 m wide and 0.30–1.00 m deep, Fig. 2) typical of the local ditch network in each of the eleven hydrological blocks. On average, the Euclidian distance to the closest studied ditch was 6.4 km (6.7 standard deviation, range 1.6 to 22.5 km). Crayfish were collected once a year over seven consecutive years (2015–2021), during the annual peak in crayfish activity, around mid-June (similar pattern to other French regions; Paillisson et al. 2012; Meineri et al. 2014). No extreme meteorological event occurred during the fieldwork itself, or the preceding few days, in particular heat waves or heavy rainfall which can impact crayfish activity. Sampling was performed using 25 unbaited funnel traps (wire minnow trap characteristics: length × width × height: 50 × 29 × 21 cm, 5.5 mm mesh size with two conical side entrances, 4 cm inner opening diameter) set approximately every 10 m along the shoreline of each ditch for a period of 24 h [a somewhat similar protocol to that used in Paillisson et al. (2012); Meineri et al. (2014)]. Using this type of trap without bait limits the risk that captures are biased by size and sex unlike other types of traps (Frutiger and Müller 2002; Paillisson et al. 2011). Besides, trapping was limited to a 24 h period because extending the trapping duration to longer periods has been found to reduce trapping efficiency per unit effort due to avoidance behaviour of crayfish towards other individuals already in the traps [by a factor of 3–4 in Paillisson et al. (2012)] and to significantly underestimate the abundance of small crayfish due to cannibalism occurring inside the traps (Paillisson et al. 2012). Thanks to all these precautions, our data are truly indicative of local crayfish (structure) populations. All individuals were collected in accordance with French regulations and preserved by freezing until further processing. In the laboratory, defrosted crayfish were basically sexed by inspection of the first pair of pleopods and measured to the nearest millimeter from the tip of the rostrum to the end of the telson (total body length) to distinguish small from large crayfish (see below and Suppl. material 1: fig. S1). Captures were then expressed in capture per unit effort (CPUE), i.e., the number of small and large crayfish per trap per 24 h. Lastly, we assumed that the number of individuals collected was too low to impact natural fluctuations of crayfish abundance over time.

Figure 2. 

Pictures of study ditches (by Aquascop).

The water regime in the ditches was described using three independent variables: i) the monthly water level (cm) in June, when crayfish were sampled, ii) the duration of the dry period over the past annual water cycle (days, over the previous 12 months), and iii) the length of the spring period (days, March to May) during which at least 20% of the grassland area adjacent to each ditch was inundated, as a proxy of the amount of potential temporary habitats for crayfish, in particular as nursery habitats for small individuals (Meineri et al. 2014). The first two variables were obtained by automatic daily monitoring of the water level in a nearby, permanent, and deeper ditch in each hydrological block, in combination with the topography of the bottom of each sampled ditch which was measured in situ using a graduated pole and was found to be stable over the study period (see values in Suppl. material 1: table S1). The water level in June correlates strongly with the monthly water level from March to May (Mauchamp et al. 2021) and was therefore considered representative of the spring water conditions. The third variable was calculated based on the principle that a pixel of 1 m² of grasslands on a LIDAR image raster was considered as flooded when its altitude on the digital elevation model of the Marais poitevin was lower than the water depth recorded in situ using piezometers (see Rapinel et al. (2018) for a complete description of calculation). Flooding duration values calculated for other grassland area quantiles (5 and 10%) were highly correlated and were therefore not retained for subsequent analyses.

Physico-chemistry variables were collected in a companion study (UNIMA, unpublished data). Water conductivity (µS cm^-1), the chlorophyll a content (mg L^-1), and dissolved organic matter (mg L^-1) were measured in a large reference ditch for each hydrological block every two months. Water conductivity was measured in the top 30 cm of the water column using a portable electronic multi-parameter probe (WTW 3430, Thermo Fisher Scientific Inc.) while chlorophyll a (measured by spectrophotometry after acetone extraction) and dissolved organic carbon contents were measured in the laboratory from a water sample collected from the top 30 cm of the water column. Average values were calculated over the February-June period each year to account for conditions during the water cycle prior to crayfish sampling (Suppl. material 1: table S1).

We also considered some habitat features expected to be relevant for crayfish diet and burrowing activities (Souty-Grosset et al. 2016). They were i) ditch bottom substrate type – determinant for crayfish (Donato et al. 2018; Palmas et al. 2019) –, and more specifically mud depth (cm) since all ditches had a muddy bottom, ii) aquatic plant cover (%) and iii) riparian vegetation cover, differentiating herbaceous vegetation from trees using scaled indices and visual evaluation [following Mauchamp et al. (2021)], all as proxies of shelter and litter input in ditches (Kobayashi et al. 2011), and iv) ditch density (m Ha^-1) in a 500-m buffer area around each ditch studied as ditches may constitute refuges or source habitats for crayfish (map source: IGN topographical database version 2008, calculation using QGIS 3.2, QGIS Development Team 2018; Suppl. material 1: table S1). Riparian vegetation was characterized on the basis of the 2015 in situ measurements as no management actions (such as pruning) have been observed over the study period.

Finally, since predatory fish may control red swamp crayfish populations (Aquiloni et al. 2010; Tablado et al. 2010; Reynolds 2011), we also used annual predatory fish biomass as a proxy of predation pressure. It was estimated for each ditch using electrofishing [data from Crabot et al. (2023)]. We took the following species into account: black bullhead (Ameiurus melas), European eel (Anguilla anguilla), pike (Esox lucius), pikeperch (Sander lucioperca), largemouth black bass (Micropterus salmoides), pumpkinseed (Lepomis gibbosus), and European perch (Perca fluviatilis).

All statistical analyses were run with R 4.3.0 (R Core Team 2023).

To distinguish small from large crayfish, we used a polymodal decomposition method that consists in fitting Gaussian components to body length-frequency data using the mixtools package (Benaglia et al. 2009). Length-frequency data showed a bimodal distribution with two distinct cohorts each year (small and large individuals, Suppl. material 1: fig. S1) with a body length threshold set at 78 mm (seven years pooled) and no difference in body size between sexes (data not shown). We tested for possible temporal autocorrelation of local crayfish CPUE (all individuals combined). In practice, we calculated correlations between CPUE in year t and year t+lag, lag ranging from 1 to 3 years in accordance with crayfish life time values (Scalici and Gherardi 2007; McLay and van den Brink 2016; van Kuijk et al. 2021) using the autocorr function in R. We also explored the relationships between the CPUE of large crayfish in a given year and the CPUE of small crayfish in the previous or the next year using linear correlations. These two complementary analyses were conducted to better understand crayfish demography and detect possible density-dependent effects.

We studied possible relationships between variation of the CPUE of the two size classes and environmental variables (Suppl. material 1: table S2) using generalised linear modelling. Dissolved organic matter in water was excluded from the suite of candidate predictors due to high multicollinearity with the chlorophyll a concentration [variance inflation factor procedure, r > 0.7 (Dormann et al. 2013), usdm package (Naimi 2017)]. All (continuous) environmental variables were scaled. Generalised linear mixed models (GLMMs) were run separately for small and large crayfish data, with a negative binomial error distribution, a log-link function, and ‘year’ as a random effect [glmer.nb function, lme4 package (Bates et al. 2015)]. We computed models from all additive combinations of up to five candidate environmental variables (and the null model) to limit model complexity [dredging procedure, MuMin package (Bartoń 2016)]. The models were ranked according to the AICc corrected for small sample sizes (Burnham and Anderson 2002), and marginal R² was provided for each model [r.squaredGLMM function in the MuMin package (Nakagawa et al. 2017)]. Conditional model-averaged estimates for each selected variable were calculated when several competitive models occurred in the model selection with ΔAICc < 2. Results were qualitatively identical over different model selections (ΔAICc < 3 and ΔAICc < 4, data not shown).

The CPUE of the two size classes varied among ditches and over time throughout the study period, but no global trend emerged (Fig. 3A, B; mean ± SD values in Suppl. material 1: table S3A, B). Small crayfish were predominant (70.7%, n = 6,272 individuals, Fig. 3A) and the sex ratio was slightly skewed (57% females). Large between-year variation in crayfish CPUE was observed in five of the 11 ditches (ditch 3, 5, 7, 10, and 11 depending on size classes). CPUE varied up to ten- and six-fold between years for small and large crayfish, respectively (Fig. 3A, B). Crayfish CPUE was sometimes very low locally; for instance, only one large crayfish was captured in ditch 1 in 2015, and none at all in ditch 5 in 2019.

Figure 3. 

Variation of red swamp crayfish in 11 ditches (one per hydrological block) expressed in capture per unit effort (CPUE, number of individuals per trap per 24 h) during the seasonal peak over time (2015 to 2021): A) small, B) large crayfish. Bold black lines represent average values for all ditches.

No temporal auto-correlation of crayfish CPUE was found in any ditch, whatever the time lag and the size class (Suppl. material 1: fig. S2A, B). Moreover, no significant correlation was found between the CPUE of large crayfish in one year and the CPUE of small ones in the previous or following year.

Variation of crayfish CPUE was consistently related to three habitat features, at both size classes: crayfish were more abundant in ditches with densely vegetated banks [both for tree hedge (estimate ± 95% CI: 0.82 ± 0.32 for small and 0.96 ± 0.28 for large crayfish) and herbaceous vegetation covers (0.39 ± 0.26 for small and 0.70 ± 0.21 for large crayfish, Fig. 4A, B)] and less abundant when the ditches were part of a dense network of ditches (-0.46 ± 0.31 and -0.85 ± 0.23 for small and large crayfish, respectively, Fig. 4A, B). In addition, variation of small crayfish CPUE was positively related to water conductivity (0.53 ± 0.33, Fig. 4A). Only one water regime descriptor was found to be significant to crayfish: dryness duration of ditches had a positive effect on the CPUE of small crayfish (0.46 ± 0.44, Fig. 4A). Aquatic plant cover and predatory fish biomass did not explain the variation in crayfish CPUE for either size class. The significant environmental variables all together explained 54 and 57% of variance of small and large crayfish CPUE, respectively (Suppl. material 1: table S4).

Figure 4. 

Forest plot of estimates (with 95% confidence intervals) of the environmental variables predicting the local CPUE of red swamp crayfish: A) small (R²_m = 0.54), B) large crayfish (R²_m = 0.57, model-averaging procedure). Values were calculated from the model selection ΔAICc < 2 (provided in Suppl. material 1: table S4). The sign of the relationship between a predictor and crayfish CPUE was ascertained when its 95% confidence interval did not include zero. No estimate was provided for some predictors because they were not in the model selection.

Crayfish CPUE varied over time, more so in specific ditches. In no case did they increase continuously over time, suggesting that the classical boom stage after introduction has already been reached (Strayer et al. 2017). Crayfish CPUE values fall within the same range as the CPUE values found in a similar French marshland colonised for 30 years (Bélouard et al. 2019 (Suppl. material 1: table S4); using the same sampling methodology). No case of local extinction has been observed in our study area (despite the absence of crayfish on one occasion in one ditch and occasional low abundance values in some ditches). All this suggests that crayfish were in the persistent phase characterised by no specific long-term abundance trend (Strayer et al. 2017). Different scenarios of long-term dynamics trends have been documented for red swamp crayfish in European regions: no trend in abundance (Sousa et al. 2013, the present study), positive trends (Soto et al. 2023), and negative trends (Bélouard et al. 2019; Soto et al. 2023). Bélouard et al. (2019) even observed a global decline in red swamp crayfish abundances in ponds in a nearby region of our study area, with few cases of local population collapse (personal communication). We suggest that the particularly high density of the ditch network in the Marais poitevin (11.9 km per km² in our specific study area) might facilitate possible local ditch recolonisation from nearby crayfish sources (see also below), since no prolonged local population collapse was recorded.

Besides, the high proportion of small crayfish reported in all ditches is in line with what is reported in the literature (Anastácio and Marques 1995; Scalici and Gherardi 2007; Paillisson et al. 2012) and reflects the high reproductive potential of the red swamp crayfish (Huner 2002). However, in some populations (e.g. Scalici et al. 2010), the occurrence of large individuals is higher, but data in the literature are often biased by the traps used which are selective for large individuals.

Contrary to our expectations, variation of CPUE of the two crayfish size classes is only weakly associated with water conditions, with only a positive effect of dryness duration of ditches on juvenile abundance. Crayfish CPUE was not higher where adjacent grasslands were inundated for long periods of time, whereas opposite findings have been reported, for instance, in Paillisson et al. (2011) and Meineri et al. (2014); temporarily flooded habitats were found attractive for small red swamp crayfish. Focusing on adults, Gavioli et al. (2018) showed that high water levels had a negative effect on the population size of red swamp crayfish and this holds true in a wide range of habitat types (from canals to flowing rivers). Here, as in most studies (Table 1), no such outcome has been found. It can be speculated that the extent of variation in water conditions was not large enough to influence crayfish abundance, in the Marais poitevin or in most of studies gathered in Table 1. We also interpreted the absence of an effect of water conditions as a consequence of the ability of crayfish to adapt to deteriorated environmental conditions (e.g. drying-up episodes) either by pulling back into burrows (Huner and Barr 1991; Nÿström 2002) or by temporarily moving to more favourable nearby habitats (Fonteneau and Paillisson 2014). Further investigations would be needed to explore these hypotheses.

Small and large crayfish were more abundant both in ditches with densely vegetated banks and in hydrological blocks composed of a less dense ditch network. Densely vegetated banks principally provide crayfish refuges (notably tree roots and woody debris jams) from predators (Holdich et al. 2006) and also food resources (input of decaying plant matter; Kobayashi et al. 2011). Regarding aquatic vegetation, it is expected to be attractive for crayfish, but negative feedback effects of red swamp crayfish on aquatic vegetation have also been extensively documented (e.g., consumption, bioturbation; Rodríguez et al. 2005; Twardochleb et al. 2013). No such relationship was found in our study, as in others (Table 1), suggesting that observed crayfish abundances were too low to significantly impact the aquatic plant cover. It may also be that dominant plant species (Elodea canadensis, E. nutalii, Stuckenia pectinata and Ceratophyllum demersum; Mauchamp et al. 2021) have the ability to efficiently regrow after being consumed by crayfish (Hessen et al. 2004); alternatively, some species are not impacted by red swamp crayfish (e.g. E. nutalii; Chucholl 2013). A third hypothesis could be that annual drying-up episodes do not allow crayfish to constitute large populations limiting their impact on aquatic plants, while, concomitantly, drying periods promote aquatic vegetation (in particular amphibious plant species) as shown by Mauchamp et al. (2021) in the same study ditches.

The negative relationship we found between crayfish CPUE and ditch density could be tentatively interpreted by a dilution effect of crayfish abundance with increasing availability of aquatic habitats. With a dense network of ditches, crayfish may also easily escape unfavourable – dry – conditions and, conversely, (re)colonise ditches. Concomitantly, the aquatic continuum between ditches is frequently disrupted by artificial impoundments and local drought (field observations), suggesting that a relationship between density and connectivity may not be straightforward. Additional investigations are clearly needed to better interpret the relationship between crayfish abundance and ditch density.

Finally, the absence of an effect of water conductivity on large crayfish was expected since the observed values were mostly below the known tolerance threshold (Kang and King 2012; Meineri et al. 2014; Palmas et al. 2019; Dörr et al. 2020). However, small crayfish were more abundant in ditches with higher water conductivity and this remains unexplained because they have a lower tolerance to water salinity than large ones (5 g L^-1 vs. 10 g L^-1; Paillisson et al. 2012; Meineri et al. 2014).

As indicated above, the first challenge of this study was to provide robust long-term data on the structure of local populations in drainage ditches using one sampling per year. The second objective was to identify environmental drivers of crayfish abundance, some being calculated on a yearly basis. For this purpose, sampling was performed during the typical annual peak in crayfish activity in spring (see also Meineri et al. 2014). At this time of year, adjacent temporary habitats have dried up and crayfish retreat into ditches, which corresponds to the best conditions for providing a good overview of local crayfish populations in ditches. Moreover, great care has been taken to limit trapping bias, although we cannot exclude potential short-term confounding factors (e.g. weather conditions) that may influence crayfish activity. In any event, the sampling protocol was consistent within each time-series which allowed us to compare spatiotemporal trends of red swamp crayfish abundance between ditches. As a result, environmental factors explained a significant part of variation in small and large crayfish CPUE over time and are thus reliable and robust. For all these reasons, the present results come from consistent datasets. A potential improvement would be to consider additional biotic factors, notably food resources to which crayfish are likely to be sensitive; however, such information is difficult to acquire explaining why they are so scarce even in short-time studies. Lastly, considering the sexual maturation status of red swamp crayfish rather than their body size could provide a valuable perspective since size-at-maturity is very variable; it can be influenced by environmental conditions (Anastácio and Marques 1995; Huner 2002) and it can, in turn, impact local population dynamics.

There is a broad consensus within the scientific community on the need to understand how invasive species are influenced by environmental conditions to predict their population dynamics (Simberloff et al. 2013). However, as reported by Soto et al. (2023), long biomonitoring data of crayfish invasions (on crayfish themselves as well as environmental data) are lacking preventing large spatio-temporal analyses of their population dynamics. The conclusions from our time-series study strengthen some of the findings derived from short-term studies. In particular, much of the variation in crayfish abundance we found was weakly explained by water conditions (temporary droughts) in drainage ditches – at least within the range of values tested in the present study –, whereas it was positively related to riparian vegetation and negatively linked to the density of surrounding habitats (ditches). Predatory fish biomass did not explain crayfish abundance. Studying growth and mortality parameters using length-frequency data collected at an increased frequency in conjunction with environmental conditions would provide additional valuable information, but it remains highly challenging when conducting time-series studies.

A direct application of our study is to provide recommendations to managers for mitigating red swamp crayfish populations, although the management of biological invasions is a complex task (Simberloff et al. 2013). Managing the riparian vegetation of ditches would be an interesting lever to reduce refuges from predators and food resources for red swamp crayfish. However, it would probably result in undesirable negative effects on other aquatic organisms, as shown in companion studies (Mauchamp et al. 2021; Crabot et al. 2022, 2023). An alternative management option would be to target management actions on small populations. Classical control actions, notably harvesting as reviewed in Freeman et al. (2010) and Gherardi et al. (2011), can significantly reduce crayfish abundance in small patchy habitats. The challenge in our study area is to mitigate crayfish abundance in connected aquatic habitats. This means that limiting crayfish dispersal using physical barriers would be required in parallel with control operations (Coignet et al. 2012; Reisinger et al. 2024). However, such management actions are very time-consuming and cannot be conducted over the long term (Strayer et al. 2017). More broadly, any management action aimed at maintaining or increasing local biodiversity should make ecosystems more resistant to invasive species. Finally, we encourage long-term studies to be carried-out on red swamp crayfish (as well as environmental and biotic factors) in a diversity of aquatic ecosystems to better understand their population dynamics (see Soto et al. (2023) in European rivers).