Citizen science dataset on the distribution of Odonata species in the area Indre-et-Loire in France

Here, our objective was to use a dragonfly citizen science database to identify both i) the spatial scale at which landscapes influence Odonates diversity in water bodies and ii) the effects of intensive agricultural landscapes on this diversity. We also identified the potential biases inherent to this particular database in order to assess to what extent it can be used reliably to infer the effects of intensive agriculture on Odonates’ species richness. The quantification of the distance up to which agricultural lands alter the diversity of Odonates is of paramount importance for (i) deciding which wetland areas are better candidates for applying efficient conservation strategies focusing on habitats (i.e., areas within this distance) or (ii) thinking about nature conservation policies at larger territorial scales. We compiled 7,731 observations made over 10 years by naturalists in a portion of the region Centre-Val-de-Loire (France) on 729 water bodies to analyse separately the effect of agricultural landscapes on the species richness of damselflies (Zygoptera) and dragonflies (Anisoptera). We focused on lentic systems because ponds and marshes are expected to be strongly influenced by local landscape characteristics, while river and stream ecosystems can also be influenced by quite more distant processes. However, we included the presence of rivers and streams near the water bodies in our analysis to control for potential dispersion of generalist species from lotic to lentic systems which could positively affect observed species richness independently of surrounding agricultural landscape. Citizen science databases are implicitly “imperfect” data because most observations are not necessarily done according to standard protocols and detection methods (Johansson et al. 2020).
Study area
Our study area is the whole Indre-et-Loire administrative department in west-central France (Fig. 1a). This department covers a surface of 6,127km2 (centroid: 47°15’29.00’’N, 0°41’29.50’’E – EPSG 4326) and altitudes ranging from 26 to 187m (mean=100m). The climate is temperate with relatively warm summers (mean monthly temperature from 11.1°C to 25.5°C) and mild winters (mean monthly temperature from 1.9°C to 12.3°C). Rainfall is relatively low, around 650mm per year (fr.climate-data.org). The landscape of the whole region consists mainly of cultivated fields (56%) and forests (24%), leaving few portions for grape or urban areas (CORINE Land Cover, 2012). This territory is particularly heterogeneous regarding aquatic habitats because it includes several large rivers and streams with quite varied water flows, and a large variety of ponds, as well as marshes and peatlands. The area includes various types of ponds including natural and artificial water bodies, ponds with (and at various levels of vegetation diversity) or without aquatic vegetation, and in open as well as closed (forestry) environments (but see below for a description of water body types).
Dragonfly distributions
More than one hundred participants, mainly members of naturalists’ associations, were involved in this biodiversity inventory project. Dragonfly presence data were georeferenced and generated directly in the field through apps for mobile devices (ObsMapp and iObs). Subsequently, the data were extracted from a local sub-site of Observation.org, a worldwide observation-reporting system that provides high quality data for all taxa (Hochachka & Fink, 2012). Each data was validated by experts as a part of a local dragonfly atlas program (https://anepe-caudalis.observation.org/). These dragonflies and damselflies data are also available through the Global Biodiversity Information Facility (https://www.gbif.org/) and are part of the French National Inventory of the Natural Heritage coordinated by French Office of Biodiversity (OFB), National Centre for Scientific Research (CNRS) and National Museum of Natural History (MNHN).
We considered information relevant to lentic hydrosystems and to contemporary landscapes by selecting observations collected inside water bodies and within a 50m radius around them between 2007 and 2017 (n=7,731 observations; an observation consists of a single species sight at specific geographic coordinates for a given day and contains number of individuals as well as reproductive or behavioural information). Hence, for each water body and each species, an index of potential autochthony was determined from indices showing that reproduction occurred (exuviae, emergence, egg laying or mating behaviours) and/or from the number of individuals in the case of damselfly species (reproduction was considered as “effective” for a species when ≥5 individuals were present). Citizen data are necessarily ‘imperfect’ data with several biases including no observation of reproduction cues for species that are actually autochthonous (e.g., survey was done within a daily window when species are not mating or laying eggs) and observation of reproduction cues for species that actually cannot develop within the water body where they lay eggs (e.g., presence of pollutants in the water). Therefore, this index of autochthony should be interpreted as a deliberately simple proxy for specifying the ecological level at which the effects of intensive agriculture may have the strongest impact. In addition, strict lotic species were automatically excluded (Table S1). Then, the data collected during the 10 years were pooled to compute the overall species richness (OSR) and the autochthonous species richness proxy (ASR) for each water body. Those two indexes were also calculated separately for dragonflies (OdragSR and AdragSR) and damselflies (OdamSR and AdamSR). Water bodies (n=729) were distributed over the entire study area (Fig. 1b) and categorized according to the habitat classification given in the French dragonflies monitoring program INVOD (Dommanget 2002) and by including some additional information on basic habitat diversity (categories are provided in Table S2). The area of water bodies ranged from 0.01 to 55.3 ha (mean 1.76 ha and median 0.29 ha; see Fig. S5-B for the complete distribution of water body areas).
Landscape characteristics
We retrieved the landscape characteristics around the water bodies using the CORINE Land Cover classification which is a pan-European land cover database. This database reports land use and cover based on images taken approximately 1–2 years before the release date (here, 2012) (Soukup et al., 2016). Around each of the 729 water bodies, we generated a set of landscape variables calculated for areas within a given distance from the water body. We generated four different landscape ‘buffers’ corresponding to areas within a distance of 200, 400, 800 and 1,600m from the water body using Q-GIS 2.14.11 (Fig. 1c). We chose the 200m landscape scale as the smallest reliable buffer given the CORINE Land Cover mapping scale (1:100,000). In accordance with previous published data on Odonata dispersal in agricultural landscapes, we chose not to go into a buffer larger than 1,600m (Conrad et al., 1999). These buffer sizes include the estimated distances travelled by non-territorial damselflies (e.g., coenagrionids) (Rouquette & Thompson, 2007) and, even if dragonflies (Anisoptera) can of course travel over larger distances, most of them are strongly territorial when sexually mature and remain relatively close to their water body. Although it is acknowledged that overlapping landscape buffers may not necessarily violate statistical independence (Zuckerberg et al. 2020), we also chose not to investigate larger radii to minimise spatial overlap between sites. For each water body and buffer, we calculated the water body area, the proportion (%) of intensive agriculture (CORINE Land Cover, 2012; Büttner et al., 2004) which represented the main land-use pressure in our study area, as well as the total length of streams and rivers (m/ha) (IGN, 2012). To ovoid collinearity issues, we did not include the proportion of forest cover in our models. Indeed, this proportion is strongly and negatively correlated with intensive agriculture in our system. In the CORINE Land Cover database (Büttner et al., 2004), we used the category of “arable lands” (code 2.1) as a proxy for intensive agriculture. In our study area, this arable land corresponds to this type of conventional agriculture that rely on frequent use of pesticides and chemical fertilizers. This intensive agriculture is locally dominated by wheat, rapeseed and sunflower (Fig. S1). In addition, the proportion of the land around water bodies covered by intensive agriculture (in a radius of 1,600m) remained similar between 2006, 2012 and 2018 (Fig. S2). Therefore, the snapshot of 2012 is representative of the intensive agriculture coverage for the whole period 2007-2017 when Odonata were observed by citizens.