Effects of Different Types of Agricultural Land Use on the Occurrence of Common Aquatic Bugs (Nepomorpha, Heteroptera) in Habitats with Slow Flowing Water in Bulgaria, Southeast Europe

Abstract

Agricultural activities can have a significant impact on aquatic organisms, including aquatic insects. Most of the aquatic Heteroptera are known as moderately tolerant to low oxygen and high nutrient concentrations. Nevertheless, the complex effects of agriculture (source of both pesticides and nutrient loads) on this group are still unclear. Therefore, the relationship between six agricultural land use classes and the occurrence of common aquatic bugs in Bulgaria was studied. In order to avoid detection bias, presence-only models were applied; Maxent algorithm was used. According to the results, land use practices connected to arable land (annual crops) have stronger influence on the occurrence of the selected aquatic Heteroptera species than those connected to perennial crops (vineyards and fruit trees). Higher sensitivity to the effects of agriculture was indicated for species preferring microhabitats without macrophyte vegetation, Aphelocheirus aestivalis (Fabricius, 1794) and Micronecta griseola Horváth, 1899, compared to species preferring macrophyte dominated sites, Nepa cinerea Linnaeus, 1758, Ilyocoris cimicoides (Linnaeus, 1758) and Sigara striata (Linnaeus, 1758).

1. Introduction

Freshwater ecosystems are exposed to rapid environmental changes and are among the most endangered types of ecosystems worldwide [1]. The adverse effects of excessive nitrogen (N) and phosphorus (P) inputs are a major issue for water quality and ecosystem health [2,3]. Agricultural areas are among the main sources of nutrients in surface waters [4,5]. Intensive agriculture has strong influence on aquatic ecosystems, since nutrient inputs increase eutrophication [6,7]. Furthermore, agricultural areas are the major source of pesticide contamination of surface waters, which has been discussed as a major water quality problem in western Europe [8]. Thus, agricultural activities could have a significant impact on aquatic organisms, including aquatic insects [9,10,11,12,13]. Some groups of aquatic insects have been reported as sensitive indicators of long-term environmental changes in water and habitat quality [14]. In particular, the Ephemeroptera, Plecoptera, and Trichoptera (EPT group) species are considered good biological indicators in stream ecosystems [15]. Most of the aquatic Heteroptera (also called water bugs) have moderate tolerance to the decline in the ecological quality of water, in respect to oxygen and nutrients concentrations [16]. Following that notion, Nepomorpha representatives at the species or genus level have rarely been singled out as potential indicators in biomonitoring systems [17]. Nevertheless, significant differences in the sensitivity to organic pollution between taxa (lower than family) belonging to Nepomorpha have been well documented [18,19,20,21,22]. Additionally, the usage of species of the family Corixidae (Nepomorpha) as indicators of organic pollution has been proposed [21], and a methodology for water quality monitoring for lakes in Finland based on the species composition of genus Micronecta (Micronectidae, Nepomorpha) has been described [18]. In respect to the sensitivity of aquatic Heteroptera to the negative effects of chemical insecticides, a recent review [17] has shown that no general conclusions can be drawn based on the available data, exception being the connection between pesticide load and decline of species number. Therefore, effects of chemical insecticides, as well as the complex effects of agricultural activities (sources of both pesticide and nutrient loads) on aquatic Heteroptera need more attention.

Land use/cover variables have been among the most important factors determining the structure of Heteroptera assemblages in Greece, Southeast Europe [23]. It can be assumed that such variables, as the proportion of the area occupied by different land use classes, might be a useful approximation for the effect of agricultural land use on the occurrence of aquatic Heteroptera in Bulgaria, Southeast Europe.

In the present study, the relationship between agricultural land use and the occurrence of common aquatic bug species was analyzed with the following aims: (1) to determine whether agricultural land use impacts the occurrence of common aquatic bugs and if so, (2) which types of agricultural land use have the strongest effect, (3) and which of the target aquatic bug specie are most sensitive, in respect to their occurrence, to the effects of agricultural land use.

2. Materials and Methods

2.1. Study Area and Species Occurrence Data

Data on the occurrence of aquatic bugs was collected for all four basin districts on the territory of Bulgaria, belonging to two ecoregions: Danube River and Black Sea Districts, belonging to the Black Sea catchment and united as Ecoregion 12 (Pontic Province); East and West Aegean River Basin Districts–Ecoregion 7 (Eastern Balkans). The basin districts were delimited according to the Water Framework Directive 2000/60/EC (WFD), the ecoregions following Cheshmedjiev et al. 2010 [24]. According to the map presented on map A of WFD 2000/60/EC (www.eea.europa.eu, accessed on 11 February 2023) the Black Sea River Basin district follows in two ecoregions: its northern part in Pontic Province and its southern part, in the Eastern Balkans. Since such separation has not been based on biogeographical data [24], for the description of the study area here the ecoregions’ boundaries proposed in Cheshmedjiev at al 2010 [24] were applied.

To control for the influence of habitat type on the occurrence of the species, the analysis was restricted to records of aquatic bugs collected from one habitat type, according to the European nature information system (EUNIS) classification–C2.3. Permanent non-tidal, smooth-flowing watercourses defined as “Permanent water courses with non-turbulent water and their associated animal and microscopic algal pelagic and benthic communities. Slow-flowing rivers, streams, brooks, rivulets and rills; also, fast-flowing rivers with laminar flow. The bed is typically composed of sand or mud.”. The EUNIS habitat classification was chosen, since it is a comprehensive pan-European system, covering types of habitats from terrestrial to freshwater and marine; from natural to artificial (https://eunis.eea.europa.eu, accessed on 11 February 2023).

Only species with occurrence data for ten or more localities of the cited habitat type in Bulgaria were included in the following analysis.

Data on the occurrence of aquatic bugs were collected from two sources: 116 records based on published data about aquatic invertebrate composition in various water bodies in Bulgaria [25,26,27,28,29,30,31], 102 records based on materials part of the “Aquatic Invertebrates Collection” maintained by the Institute of Biodiversity and Ecosystem Research at the Bulgarian Academy of Sciences (IBER-BAS). The records were connected to 177 hydrobiological samples taken from 139 localities (Figure 1). Samples were collected between 2005 and 2020. Most of the materials were collected using the adapted version of the multi-habitat sampling method [32], according to the accepted standard procedures (EN 16150: 2012, EN ISO 10870: 2012).

2.2. Environmental Dataset

A GIS layer consisting of river catchments in Bulgaria within Ecoregions 12 and 7, and including each catchment’s respective area, minimal altitude, minimal slope, and identification code of the downstream adjacent catchment (sub-basin) was extracted from the CCM2 geodatabase [33]. This dataset is a subset of Catchment Characterization and Modelling (CCM) data version 2.1 provided by EU Joint Research Centre (JRC). The CCM2 database covers the entire European continent and includes a hierarchical set of river catchments and segments based on the Strahler order and structured hydrological feature codes based on the Pfafstetter system [33]. The Heteroptera occurrence data were added to the catchments layer attributes. The further steps of the data preparation included only catchments with minimal altitude and minimal slope, both less than or equal to their maximums among the catchments with Heteroptera records. This was done in order to ensure that only catchments connected to river sections with characteristics relatively close to these of smooth flowing water courses (such as those with Heteroptera records) would be used in the analysis.

Corine land cover (CLC) data was derived from three geodatabases (in vector format) downloaded from the official cite of Copernicus—The European Earth Observation Programme [34]: (1) containing CLC 2006 data (U2012_CLC2006_V2020_20u1.gdb); (2) containing CLC 2012 (U2018_CLC2012_V2020_20u1.gdb); (3) containing CLC 2018 (U2018_CLC2018_V2020_20u1.gdb). In order to extract the needed land cover data, the following steps were caried out using QGIS 3.16 [35]. The CLC layers as well as the catchment layer were reprojected to coordinate reference system 32635–WGS84/UTM Zone 35N, for more accurate area calculations. Then each of the CLC layers was separately combined with the catchment layer by applying the geoprocessing function “Intersection”. The resulting layers consisted of polygons each of which belonged to one land use class and to one river catchment, respectively. This enabled the calculation of the cumulative area of each agricultural land cover third level classes (

Table 1. Corine land cover nomenclature for Agricultural areas [34].

LEVEL 1	LEVEL 2	LEVEL 3
2. Agricultural areas	2.1. Arable land	2.1.1. Non-irrigated arable land
	2.2. Permanent crops	2.2.1. Vineyards
		2.2.2. Fruit trees and berry plantations
	2.3. Pastures	2.3.1. Pastures
	2.4. Heterogeneous agricultural areas	2.4.2. Complex cultivation patterns
		2.4.3. Land principally occupied by agriculture, with significant areas of natural vegetation

), as the area of a given catchment plus the areas of its tributaries’ catchments and/or that of immediately upstream catchment covered by a given land cover class. The cumulative area of each agricultural land cover class was divided by the sum of the total area of the relevant catchments and then converted to percentage. Such variables (cumulative area of land cover classes) have been used in previous studies on the ecology of aquatic invertebrates on catchment scales [36,37], aquatic Heteroptera in particular [30]. The agricultural variables obtained for each of the three time periods were tested for correlation using the built-in function cor in R version 4.2.2 [38] for calculating the Kendall rank correlation coefficient. The results showed that the agricultural variables for CLC 2006 and CLC 2018 were strongly correlated with those for CLC 2012 (Kendall’s Tau correlation coefficients 0.96–0.98, and 0.84–0.95, respectively, with p < 0.000001). Although the land cover in the studied catchments undoubtedly has changed during the sampling period (2005–2020), the proportion of their area covered with the target agricultural land cover classes seams stable enough for the purpose of the present study and the analysis of multi-year species occurrence data it involves. In further steps, the agricultural variables for CLC 2012 were used, since this period is closer to the middle of the sampling period.

The data obtained for the analysis (including the CLC 2006 variables) are available in Table S1, and the related metadate in File S1.

2.3. Modeling the Response of the Selected Species to the Agricultural Land Use Variables

Prior to the modeling, the correlation among the catchment’s variables (area, minimal altitude, minimal slope and the agricultural variables was evaluated, using the built-in function cor in R version 4.2.2 [38] for calculating Spearman’s rank correlation coefficient.

Predictions of the occurrence probability for each of the nine aquatic bug species on catchments scale were built, based on each of the target land cover variables used individually by Maxent version 3.4.3 [39] and following the method described by Elith et al. (2011) [40], with settings as outlined in the next paragraphs. Maxent software requires input of environmental data for all locations (here scaled to catchment) in the studied area. Our choice for applying this algorithm was based on its good performance when used for presence-only models [40,41,42,43]. A presence-only approach was selected in order to avoid detection bias—smaller species, for example, those of genera Micronecta Kirkaldy, 1897 and Plea Leach, 1818, may go unnoticed in a given locality much easier than more conspicuous water bugs (such as Nepa Linnaeus, 1758 and Ranatra Fabricius, 1790).

The input for the analyzed species were spreadsheet-like summaries of environmental variables in two comma separated value (CSV) files (“Sample file” and “Background file”). The “Sample file” included all of the catchments with records of the respective species, while the “Background file” included all catchments with minimal altitude and minimal slope, both less than or equal to their maximums among the catchments with records of the respective species, and also area of the catchment greater than or equal to its minimum in the “Sample file”. The described restrictions to the background data were applied in order to control for these three non-target parameters, each of which might exert strong influence on the occurrence probability of the analyzed aquatic Heteroptera, masking the effects of the target agricultural variables.

All of the environmental variables were continuous. Following Elith et al. (2011) [40], the default settings for features and regularization were used and the output was set to be logistic. For obtaining out-of-sample estimates of predictive performance and estimates of uncertainty around fitted functions, a ten-fold cross-validation was used. The maximum number of background points for each of the species was set to the respective number of catchments in its “Background file”, instead of the default 10,000. The algorithm was run ten times for each species-variable combination. To measure the amount of variation in the response curves of the single variable models, for each of the species-variable combination standard deviation over the ten respective models (runs) was calculated.

Model performance was evaluated based on the area under the curve (AUC) of a receiver operating characteristic plot [44]. According to Elith et al. (2006) [41]: “AUC values can be interpreted as indicating the probability that, when a presence site and an absence site are drawn at random from the population, the first will have a higher predicted value than the second”. The value of AUC could be used for assessing the ability of maxent models’ predictions to discriminate presence from background locations (catchments in our case). For example, AUC value < 0.5 shows that the given model performs in predicting the species occurrence worse than a random model would [41]. In contrast, models with AUC values greater than 0.9 could be considered excellent in predicting a species’ occurrence based on the given data.

3. Results

The relationship between six agricultural land use variables and the occurrence of nine common aquatic bug species was analyzed. The selected species represent seven Heteroptera families: Aphelochiridae—Aphelocheirus aestivalis (Fabricius, 1794); Noucoridae Ilyocoris cimicoides (Linnaeus, 1758); Nepidae—Nepa cinerea Linnaeus, 1758; Micronectidae—Micronecta griseola Horváth, 1899; Corixidae—Sigara iactans Jansson, 1983, Sigara lateralis (Leach, 1817) and Sigara striata (Linnaeus, 1758); Notonectidae—Notonecta glauca (Linnaeus, 1758); Pleidae—Plea minitissima Leach, 1817.

Among the best models (averaged over the ten runs) for each of the species the AUC varied between 0.52 and 0.74 (Figure 2). Among the 54 created models there were 21 with noticeably high standard deviation—75% of the predicted probability values were with standard deviation (over the ten models) above 0.1 (Figure 3). Such models, as well as those with AUC lower than 0.6 (Figure 2), were excluded from further comments on the effect of land cover type on the occurrence of the selected aquatic Heteroptera species.

Pasture area was the most informative (compared to the other five variables) in respect to the occurrences of Aphelocheirus aestivalis and Sigara iactans, while non-irrigated arable land (Figure 2 and Figure 3) of Ilyocoris cimicoides and Micronecta griseola. Land principally occupied by agriculture was the best predictor for Nepa cinerea and Sigara striata.

The models created for Notonecta glauca, Plea minutissima and Sigara lateralis had either low AUC or high standard deviation, therefore, the response curves connected to these models were not commented on below.

There was an increase in the respective probability of occurrence for each of the two species, Ilyocoris cimicoides and Nepa cinerea, with the increase of the area occupied by non-irrigated arable land (Figure 4A), and only for Nepa cinerea with the increase of the proportion of the land principally occupied by agriculture (Figure 4F).

The model for Sigara iactans showed an increase in probability with the increase of the area occupied by pastures (Figure 4D). The occurrence probability of Sigara striata peaked between 38% and 69%, and at 11% in connection with non-irrigated arable land and land principally occupied by agriculture (Figure 4A,F), respectively. The probability of occurrence of Aphelocheirus aestivalis was influenced negatively by each of the four agricultural variables when their values were higher than certain levels: non-irrigated arable land—above 60%; pastures—above 5%; complex cultivation patterns (Figure 4E)—above 2%; land principally occupied by agriculture—above 9%.

In respect to the non-irrigated arable land, the model for Micronecta griseola showed a decline of the occurrence probability with the increase of the cumulative percent (Figure 4A).

The six agricultural land use variables did not show strong intercorrelations, nor correlations with the other three variables in the dataset (area, minimal altitude and minimal slope); correlation coefficients were below 0.7.

4. Discussion

The results indicated influence on the occurrence of common aquatic bugs in connection to the percent of the area occupied by four of the six agricultural land use classes: non-irrigated arable land, pastures, complex cultivation patterns and land principally occupied by agriculture. Nevertheless, among these four agricultural variables, the cumulative area of the non-irrigated arable land seems to be the one with the most prominent effect on the distribution of the selected aquatic Heteroptera, since our results indicate its influence on the occurrence of four of the nine species. Such results could be expected, as this type of land use has been connected to higher yearly loads (higher export coefficient) of phosphorus in comparison with the other three classes mentioned above [45]. Furthermore, larger areas occupied by arable land in a given stream’s catchment has been linked to higher nitrogen compound concentrations in the waters of the given stream [46]. In the present study, the area occupied by vineyards and by fruit trees and berries plantations were the two variables with the least importance as predictors of the analyzed species. This outcome might be explained by the lower significance of these two types of land use as sources of nutrients for the aquatic habitats, suggested by the results of Łaszewski et al., 2022 [46]. Additionally, the most frequently detected pesticides in waters of vineyard regions are not insecticides but fungicides, according to a study in Spain [47]. Generally, perennial crops, pastures, and trees might be more sustainable agricultural productions than annual crops [48].

The tendencies indicated by the response curves should be treated with caution, as each of the models they represent is based on one variable but shows not only the dependence of the predicted occurrence probability on that given variable, but also the dependencies induced by correlations between the given variable and other variables [49]. The correlation analysis carried out not only for the land use variables but also for those variables and three other catchment variables (area, minimal altitude, minimal slope), indicated low correlation (coefficients below 0.7). It could be inferred that the tendencies shown by the response curves (presented above) bear some relevance to the relationships between the analyzed agricultural variables and the selected aquatic heteropteran species.

Pastures might act as biogeochemical barriers due to the permanence of vegetation cover and soil rich in organic matter [46,50,51], but their effect on water quality may vary greatly, due to differences in their management and agricultural practices (fertilizing, mowing, and grazing), which could induce significant impact on water quality [52,53]. Therefore, models involving the pasture variable are treated as ambiguous and are not commented on in the sense of species sensitivity. The results concerning the rest of the analyzed land use variables suggest that sensitivity increases in the following order: Nepa cinerea (positively connected to two agricultural land use variables) and Ilyocoris cimicoides (positively connected to one agricultural land use variable) < Sigara striata (negative connection with higher values of two agricultural land use variables) < Aphelocheirus aestivalis (negative relation with higher values of three agricultural land use variables) and Micronecta griseola (negatively connected to an agricultural land use variable). This grouping coincides with the level of preference of these aquatic insects to macrophyte vegetation. The three less sensitive species Nepa cinerea, Ilyocoris cimicoides and Sigara striata, all have preference for macrophyte dominated microhabitats [54]. In contrast, the two species with highest sensitivity to the effects of agricultural land use (present study), are more closely connected to microhabitats characterized with the absence of macrophytes—sand or fine to medium-sized gravel substratum [54]. Furthermore, aquatic and riparian vegetation have been among the main factors influencing the structure of aquatic Heteroptera assemblages in other regions [23,55,56,57,58]. Therefore, it could be assumed that the main effects of the diffuse pollution from agricultural land use on the selected species are most likely connected to changes in the macrophyte cover in the given microhabitats. Nevertheless, the direct influence of pollution should not be excluded, especially in the case of Aphelocheirus aestivalis, which has been often listed as an indicator species sensitive to organic pollution (used, for example, in Germany, Austria, Czech Republic, Slovakia) based on the categorizations in [54], and in EU-STAR (2005) [59]. In Bulgaria, according to the broadly applied adapted version of the Biotic index (as described in Cheshmedjiev et al. (2013) [60]), Aphelocheirus aestivalis belongs to the second most sensitive group of macroinvertebrates, while the rest of the aquatic Heteroptera are assigned to the third group (out of five)— “relatively tolerant” taxa. The higher sensitivity of Aphelocheirus aestivalis to anthropogenic pressure is indicated in the present study as well, at least compared to Nepa cinerea, Ilyocoris cimicoides and Sigara striata. In some European countries, Aphelocheirus aestivalis has been included in local and national Red Lists or Red Data Books as endangered species [61]. Pollution from industrial and agricultural sources and watercourses regulation are pointed at as the main threats for the subpopulations [30,62] of this species. The higher sensitivity of Aphelocheirus aestivalis to polluted waters (characterized by lower oxygen concentrations) could be explained by its mode of respiration. The species is a plastron breather, meaning it relies only on diffused oxygen in the water, while most of the other representatives of Nepomorpha have bimodal breathing with gas bubbles carried on their body surface functioning both as gas stores and physical gills [63]. Additional implications for the conservation of Aphelocheirus aestivalis come from its low dispersion ability [61] and low genetic diversity [64]. Taking these facts into account, the results of the present study concerning this species might be of help for its conservation.

Interestingly, besides Aphelocheirus aestivalis, one more species showed prominent negative response to arable land use, Micronecta griseola. In different parts of its range, Micronecta griseola has been discussed both as tolerant to pollution and as a relatively sensitive species, respectively. In Finland, the species has been included in a proposed methodology for water quality monitoring of lakes based on the species composition of genus Micronecta [18]; the dominance of Micronecta griseola (together with Micronecta minutissima) has been indicative of strongly eutrophicated waters. In contrast, in regions with milder climatic conditions—Poland, Germany, Hungary—Micronecta griseola has been commented as a species sensitive to water pollution [22,65,66,67]. The differences in the tolerance of this species to polluted waters (often with low concentration of dissolved oxygen) observed between “colder” and “warmer” parts of its range might be connected to the hypothesis that the heat tolerance of aquatic insects is strongly influenced by the availability of oxygen [68,69,70,71]. Therefore, the sensitivity of Micronecta griseola to the combination of pollution and high ambient temperatures needs further investigation, especially in the conditions of climate changes and because the species has been listed as vulnerable in other regions [72]. A recent review recommends the use of Micronecta species composition for biomonitoring purposes [17]. The results of the present study confirm the need for reassessment of genus Micronecta as potential indicators of water quality. It should be mentioned that Micronecta species (especially larvae and females) are difficult to identify due to their small size (about 2 mm) and similar morphology. The high potential of DNA barcoding as a tool for the identification of aquatic Heteroptera has been demonstrated in a large-scale study in Germany and adjacent regions [73]. Further application of this technology might solve the identification problem occurring with genus Micronecta. Species of this genus, although very small in size, can produce distinct sounds by stridulation [74]. For example, Micronecta griseola has been reported to generate a very conspicuous high-pitch sound, detectable at distances of a few meters [74]. The production of such sounds (with characteristics differing between species) could enable the use of Micronecta species in ecoacoustics studies. An approach involving underwater acoustic monitoring could become a non-invasive alternative to hydrobiological sampling for monitoring of freshwater ecosystems [74], especially in localities where hydrobiological sampling is not recommended (for example, protected areas).

In the present study, the models built for three of the target species (Notonecta glauca, Plea minutissima and Sigara lateralis) had very low accuracy and/or high variation. One explanation could be that agricultural land use has negligible effect on the occurrence of these three species. Such an interpretation is in line with previous observations on the biology and ecology of these aquatic bugs. Two of them—Notonecta glauca and Sigara lateralis, have very good ability to disperse by flight [75] and high rates of migration might be masking the effect of the anthropogenic pressures on these species. Furthermore, Sigara lateralis has been found in polluted waters [76,77] and is among the few species of Corixidae occurring in eutrophic habitats even with muddy waters [75]. Both Notonecta glauca and Plea minutissima prefer macrophyte dominated habitats [75], and Plea minutissima has been often found in water bodies characterized, in summer, by low levels of dissolved oxygen [78]. Furthermore, the resilience and ferociousness of Notonectidae and Pleaidae species make them important biological mosquito control agents [79].

5. Conclusions

The results of the present study indicate that the land use variables connected to arable land (and annual crops) have a stronger influence on the occurrence of the selected aquatic Heteroptera species than variables connected to perennial crops. Species with preference to microhabitats without macrophyte vegetation can be more sensitive (in respect, at least, to their occurrence) to the effect of the agricultural activities on the ecological quality of the water than species preferring macrophyte dominated microhabitats. Therefore, agricultural activities influence the occurrence of the selected aquatic Heteroptera, most likely by inducing changes in riverine microhabitats. The present study was limited to the analysis of occurrence data and land cover data, due to the unavailability of data on other relevant variables (such as pH, conductivity, nitrogen and phosphorus compound concentrations). The main findings presented here are preliminary in nature but could help the initiation of detailed studies on aquatic Heteroptera species in respect to their sensitivity to anthropogenic pressure in Bulgaria, as well as in other regions of Southeast Europe. The use of physical as well as chemical variables in further studies on the topic is necessary. Focused research on the ecology of Micronectidae species is much needed.