Ganatra et al. Environ Sci Eur (2021) 33:58 https://doi.org/10.1186/s12302-021-00497-9

RESEARCH Open Access

Calibration of the ­SPEARpesticides bioindicator for cost‑efective pesticide monitoring in East African streams Akbar A. Ganatra1,2, Faith Jebiwot Kandie1,3,4,5, Ulrike Fillinger1, Francis McOdimba1,2, Baldwyn Torto1, Werner Brack3,5, Matthias Liess6,7* , Henner Hollert7 and Jeremias M. Becker6,7

Abstract Background: Pesticides are washed from agricultural felds into adjacent streams, where even short-term exposure causes long-term ecological damage. Detecting pesticide pollution in streams thus requires the expensive monitor- ing of peak concentrations during run-of events. Alternatively, exposure and ecological efects can be assessed using the ­SPEARpesticides bioindicator that quantifes pesticide-related changes in the macroinvertebrate community compo- sition. ­SPEARpesticides has been developed in Central Europe and validated in other parts of Europe, Australia and South America; here we investigated its performance in East African streams.

Results: With minimal adaptations of the SPEAR­ pesticdes index, we successfully characterized pesticide pollution in 13 streams located in Western Kenya. The East African ­SPEARpesticides index correlated well with the overall toxicity of 30 pesticides (maximum toxic unit maximum environmental vs. median lethal concentration) measured in stream 2 = water (R 0.53). Similarly, the ­SPEARpesticides index correlated with the risk of surface run-of from agricultural felds (as identifed= based on ground slope in the catchment area and the width of protective riparian strips, R2 0.45). Unlike = other bioindicators designed to indicate general water pollution, ­SPEARpesticides was independent of organic pollution and highly specifc to pesticides. In 23% of the streams, pesticides exceeded concentrations considered environmen- tally safe based on European frst tiered risk assessment. Conclusions: Increasing contamination was associated with considerable changes in the macroinvertebrate community composition. We conclude that pesticides need to be better regulated also in developing countries. ­SPEARpesticides provides a straightforward and cost-efcient tool for the required monitoring of pesticide exposure in small to medium streams. Keywords: Ecotoxicology, Bio-indicator, Pesticide pollution

Background and birds [2–4]. Tere is increasing evidence that the In 2020, the worldwide application of agricultural pes- pesticide-driven impairment of biocenoses also afects ticides is expected to increase from 2 million tonnes valuable ecosystem services ranging from pollination to 3.5 million tonnes annually [1]. Pesticide pollution [5] to leaf-litter degradation [6] and to the biological is considered one of the main drivers for the global control of agricultural pests [7, 8] and of pathogens decline in the abundance and diversity of insects, plants in freshwater [9]. For the United States and the Euro- pean Union, where pesticides are used on a large scale, *Correspondence: [email protected] extensive literature on exposure and efects in the envi- 6 Department of System‑Ecotoxicology, Helmholtz Centre ronment is available from academic research and from for Environmental Research GmbH – UFZ, 04318 Leipzig, Germany regulatory risk assessment [1]. In developing countries, Full list of author information is available at the end of the article

© The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​ iveco​ ​ mmons.org/​ licen​ ses/​ by/4.​ 0/​ . Ganatra et al. Environ Sci Eur (2021) 33:58 Page 2 of 15

regulation of plant protection products is often poor pollution in small to medium streams that may be spe- and information on pesticide pollution is scarce, cifcally valuable in developing countries. though recent evidence suggests that pesticide usage SPEARpesticides has been developed for temperate has been increasing [1]. As agricultural producers grow streams in Central Europe. Tus, the studied species more conscious about the use of synthetic pesticides composition, pesticide exposure patterns and ecologi- and their toxic efects [10], more efort in the assess- cal conditions may difer from those in other continents ment and mitigation of pesticide pollution is urgently such as in sub-Saharan Africa. Malherbe et al. [22] found needed. only a non-signifcant response of SPEAR­ pesticides values Freshwater macroinvertebrates in small streams are at to pesticide pollution in South African streams. In con- particular risk, since streams collect pesticide loads from trast, ­SPEARpesticides has been successfully applied and agricultural felds in the catchment area [11]. Moreover, validated to assess pesticide pollution in various streams many freshwater arthropods are highly sensitive to insec- ranging from Southern to Northern Europe [20], and ticides and fungicides [12–14]. However, the detection (after minor modifcation) in Australia [23] and Argen- and quantifcation of pesticides in streams is challeng- tina [24]. Te ­SPEARpesticides concept is based on ecotoxi- ing, as exposure occurs typically in short pulses due to cological traits rather than taxonomic relations and thus spray drift and particularly due to surface run-of from provides a mechanistic linkage of pesticide stress and the agricultural felds following heavy rainfall [15]. Such community response; in contrast to classical taxonomy- short-term exposure peaks (in the range of hours) drive based bioindicators, trait-based approaches can over- long-term efects on the macroinvertebrate community come issues with natural taxonomic variability and may for months [16–18]. Terefore, pesticide measurements thus be applicable even across diferent climatic regions from grab samples of stream water and sediment at ran- [25]. For freshwater macroinvertebrates in the South- dom time points tend to considerably underestimate the ern hemisphere, information on ecotoxicological traits magnitude of pesticide exposure. Realistic environmental is very scarce [22, 23]. However, Wang et al. [26] found monitoring in streams must capture the exposure peaks that related saltwater invertebrates from temperate and by run-of event-triggered sampling [15] or by continu- tropical regions difer only slightly in their acute sensitiv- ous passive sampling over extended periods of time [11]. ity to toxicants. Moreover, SPEAR­ pesticides can be applied However, such studies are labor-intensive and expensive at the family taxonomic level [27], and palearctic and [19]. In addition to the chemical analysis of samples for afrotropical streams share most of their macroinverte- many compounds, samplers need to be installed, pro- brate families [28]. Assuming that families of freshwater tected and regularly accessed in remote areas. Tese macroinvertebrates from temperate and tropical streams challenges limit the feasibility of monitoring pesticide may generally share their vulnerability to pesticides, pollution based on chemical analyses, particularly in SPEAR­ pesticides may be, therefore, used also in sub-Saha- developing countries. ran Africa. As an alternative approach, pesticide exposure can Here we applied the SPEAR­ pesticides bioindicator to be indirectly derived from its observed efects on macroinvertebrate samples from 13 small to medium the macroinvertebrate community composition. Te streams in Western Kenya. Te ­SPEARpesticides values SPEAR­ pesticides (“SPEcies At Risk”) bioindicator has been were compared to the toxic pressure of 30 pesticides in developed to quantify decreases in the proportion of water samples concurrently collected during the rainy those taxa considered to be vulnerable to pesticides, as season [29]. Additionally, we tested the specifcity of the compared to reference conditions [20, 21]. For this task, SPEAR­ pesticides index for pesticide efects and compared stream macroinvertebrates have been classifed as vul- its performance to alternative bioindicators for environ- nerable or non-vulnerable taxa based on ecotoxicologi- mental stressors. cal traits (see methods). Because the SPEAR­ pesticides index describes the proportion of vulnerable taxa (weighted by Materials and methods individual number), it has no unit and does not facili- Study area tate the identifcation of individual toxic compounds in Te Lake Victoria South Basin in the western part of the environment. However, ­SPEARpesticides values can be Kenya (Fig. 1) is characterized by a tropical climate with translated to an estimated toxic pressure for macroin- a major rainy season between March and June and a vertebrates in the more informative form of toxic units minor rainy season between October and December. Te (see methods). Since the ­SPEARpesticides value is driven by rainy seasons are separated by dry months, particularly long-term efects, it can be derived from a single commu- in the lowlands close to the shore of Lake Victoria. Te nity sample per site. SPEAR­ pesticides thus ofers a potential area is densely populated, but poorly developed in terms cost-efective technique for the monitoring of pesticide of infrastructure and sanitation. It is dominated by food Ganatra et al. Environ Sci Eur (2021) 33:58 Page 3 of 15

Fig. 1 Location of the sampling sites in the study area of Western Kenya, East Africa. Site locations are displayed with the site code and colour-flled

to annotate the relationship between predicted pesticide pollution (TU­ max) using ­SPEARpesticides and measured pollution in water samples using LC-HRMS in Kandie et al. [29]. Akbar Ganatra. “SPEARpesticides study sites”. “World Topographical map” & “World Hillshade”. February 2nd, 2021. https://​ arcg.​is/​1fDn9

production, mainly maize, for local markets with small before aliquots of 1 mL were transported to the labora- scale felds that often range close to the unfortifed banks tory in amber glass vials at −4 °C. Chemical analysis of streams. However, large scale commercial farming also was performed by directly injecting 100 µL of the water exists in the form of irrigated rice felds, sugarcane and sample into a high performance liquid chromatography particularly tea plantations [30]. During the rainy sea- system (Termo Ultimate 3000 LC) coupled to a high sons, streams turn red from silt loads indicating heavy resolution mass spectrometer (QExactive Plus, Termo). erosion and a high potential for surface run-of from Te water samples were subjected to target screen- agricultural felds. ing for 428 chemicals and to suspect screening for 233 additional substances. Tis analysis included 162 hydro- Pesticide measurements philic pesticides (active substances and metabolites), but Te sampling and analysis of pesticide residues has been no highly hydrophobic compounds such as pyrethroid described in detail in Kandie et al. [29]. In brief, 48 sites insecticides for which diferent sampling techniques are covering small to large streams, oxbow lakes, irrigation required. Raw data was processed using MZmine (Ver- channels, reservoirs and rice felds were sampled during sion 2.38) [31] and further confrmation and quantifca- the rainy season in September–October 2017. A single tion done using TraceFinder 4.1 (Termo) as detailed in 500 mL water sample per site was collected in a pre- [29]. Physicochemical water parameters were measured cleaned glass. Solids were allowed to settle for 1 min, in situ (fow velocity, temperature, dissolved oxygen, Ganatra et al. Environ Sci Eur (2021) 33:58 Page 4 of 15

phosphates, pH, turbidity) and from additional water The SPEARpesticides concept samples in the laboratory (nitrate, nitrite, ammonium, Te ­SPEARpesticides bioindicator has been developed to carbonate hardness). specifcally quantify pesticide-driven deviations in the In the present study, we only considered pesticides community structure from those observed under non- and converted the observed concentrations in the water polluted reference conditions. Since its frst publication samples to toxic units to quantify the toxic pressure for in Liess and von der Ohe [21], ­SPEARpesticides has been freshwater macroinvertebrates. Te toxic unit (TU) [32] constantly refned [20, 27]. Here we refer to the latest describes the environmental concentration conc of a pes- version 2019.11 that includes some major improvements ticide i in relation to its median efective or lethal con- described in Knillmann et al. [37]. centration (EC50 or LC50, concentration that afects or SPEARpesticides is based on a classifcation of freshwater kills 50% of individuals in an acute toxicity test): macroinvertebrates in” SPEcies At Risk” and in taxa being conci not at risk based on four traits that describe their vulner- TUi = ability to pesticides: (1) the average physiological sensi- LC50i (1) tivity to various pesticides (s value, mean of ­log10(LC50 For the TU calculation we generally referred to the relative to the ­LC50 of the reference species D. magna or ­LC50 for the water fea Daphnia magna after constant Chironomus sp.)); (2) the ability of autochthonous popu- exposure for 48 h as reported in the Pesticides Proper- lation recovery from reproduction (generation time); (3) ties Data Base [33]. D. magna was selected as a reference, the ability of allochthonous recovery from recoloniza- because it belongs to the more sensitive freshwater mac- tion (dispersal from non-polluted refuge areas); and (4) roinvertebrates and because most toxicological data are the probability of being actually exposed (e. g. aquatic life available for this species due to its use in regulatory risk stages during the main insecticide application season). assessment [34]. However, for some pesticides (particu- Taxa with high sensitivity, low ability of autochthonous larly neonicotinoid insecticides), D. magna turned out and allochthonous recovery and high probability of being to be considerably more tolerant than other macroin- exposed have been classifed as being at risk, others as vertebrates [34]. Terefore, we additionally extracted the not at risk. A trait data base for Central European taxa, ­LC50 for the saltwater shrimp Americamysis bahia after linked to a program for the calculation of ­SPEARpesticides 96 h exposure from the PPDB data base in case it was values is available with the software INDICATE (https://​ available. In cases, where D. magna was highly insensi- www.​syste​mecol​ogy.​de/​indic​ate). Te data base com- tive ­(LC50 > 10 mg/L), we additionally took data from the prises entries for individual species, but also for higher ECOTOX data base [35] for the standard test species taxa, where available data on lower taxonomic levels have Chironomus riparius and Hyalella azteca. If the ­LC50 of been averaged. D. magna was > 10 times higher than that of an additional Te ­SPEARpesticides index relates to the proportion of species observed after 48 h or > 100 times higher after taxa at risk within the macroinvertebrate community and 96 h, we used the most sensitive additional species as ref- decreases with increasing toxic pressure [37]: erence (mean LC­ 50 from diferent studies in ECOTOX, if n log (4x + 1) · y SPEAR = i=1 10 i available). For metabolites we used the LC­ 50 of the paren- pesticides n (2) = log (4xi + 1) tal compound if no toxicity data was available or if the i 1 10 metabolite was less toxic than the parental compound. with xi being the observed number of individuals of taxon Te overall toxic pressure from pesticide mixtures in i, and y being 1 for those taxa classifed at risk and 0 for the water samples was summarized as the summed up taxa not at risk. In this formula, SPEAR­ pesticides values can ­(TUsum, [32] and as the maximum (TU­ max, [21]) toxic range between 0 (no species at risk, indicating high pes- unit. ­TUmax and TU­ sum were set to a minimum threshold –4 ticide efects) and 1 (only species at risk). Abundant taxa of ­10 for sites at which no potentially relevant concen- are down weighted to limit the infuence of populations trations were detected; this threshold is in the range of with mass development so that the ­SPEARpesticides index the limit of quantifcation for the most toxic pesticides increases rather equally with both the incidence and the in the chemical analysis (see Additional fle 2: Table S6). population density of vulnerable taxa. To facilitate inter- Additionally, no efects of pesticides on freshwater pretation, since version 2019.11, “raw” SPEAR­ pesticides val- macroinvertebrate communities have been generally ues obtained from the equation above are divided by the observed in the feld below a threshold range of TU­ max –3 –4 average ­SPEARpesticides value observed in non-polluted between ­10 and ­10 [36]. –4 reference sites (at measured ­TUmax < ­10 ) from Germany [37]: Ganatra et al. Environ Sci Eur (2021) 33:58 Page 5 of 15

rawSPEARpesticides Te BMWP score system was developed in the 1980s ScaledSPEARpesticides = referenceSPEARpesticides and refned later to assess organic pollution in British (3) streams [40]. For a refned assessment according to the European Water Frame Work Directive, ecological qual- With reference SPEAR 0.27. Tis scaled pesticides = ity indices are derived from the BMWP score by compar- SPEAR­ index ranges from 0 (no species at risk) pesticides ing observed vs. expected values based on an elaborate to 3.7 (only species at risk) and indicates no toxic pres- stream classifcation system. Such a classifcation system sure at values > 1. For European streams, an empirical is not available for Kenyan streams; therefore, we sim- relationship between ­SPEAR and the measured pesticides ply calculated the average score per taxon (ASPT) using ­log transformed ­TU during run-of events has been 10- max the software ASTERICS–AQUEM/STAR Ecological established to convert ­SPEAR values to the esti- pesticides River Classifcation System v. 4.0.4, (https://​www.​gewae​ mated toxic pressure [20, 37]. Because the community sser-​bewer​tung.​de/​index.​php?​artic​le_​id ​419&​clang 0). composition of Kenyan freshwater macroinvertebrates = = SASS5 was originally developed in 1994 and refned in may difer from those in Germany, both the classifcation 2002 to determine the condition or ‘health’ of rivers in of individual taxa and the conversion to ­TU were sub- max South Africa [38]. As recommended, we calculated the jected to revision in the present study (see results). Average Score Per Taxa (ASPT) using the score tables available in Dickens and Graham [38]. Application of SPEARpesticides in the study area Finally, we calculated the proportion of ephemerop- We applied ­SPEAR to macroinvertebrate sam- pesticides teran, plecopteran and trichopteran insects on the overall ples collected together with the water samples from individual number (EPT) and the species diversity (Shan- Kandie et al. [29, see above]. Te macroinvertebrate sam- non index) as more general descriptors of the freshwater pling has been described in detail in Becker et al. [9]. In macroinvertebrate community. Both descriptors are con- brief, macroinvertebrates were collected at the same site sidered to decrease with increasing levels of water pollu- and day as the water samples, following a standardised tion [37]. approach adapted from the South African Scoring Sys- Additionally we estimated the run-of potential of tem 5 (SASS5, [38]. At each stream site, a 50 m stretch a site, i. e. a predictor for the risk of pesticide exposure was sampled in four quadrants by two persons in paral- from surface run-of [41]. We used a highly simplifed lel. Each quadrant was sampled for 7 min using sweep approach by classifying sites on an ordinal scale based on nets for the water surface and littoral habitats, and kick the average local ground slope and the average width of sampling for the benthic macroinvertebrates. Addition- riparian strips up to 2 km upstream as visible from online ally, gravel, soil and mud (GSM) habitats were sampled satellite imagery. Slope was classifed as fat, low, medium for 1 min per quadrant. Organisms were preserved in and high. Te run-of potential was then classifed as fol- 70% ethanol and identifed to the family level under a lows: 1 (none): 30 m bufer strip, or 20 m bufer strip dissecting microscope in the laboratory. ­SPEAR ≥ ≥ pesticides and fat or low slope; 2 (low): 21–30 m bufer strip, or v. 2019.11 and the associated estimated TU­ were cal- max 11–20 m bufer strip and fat or low slope; 3 (moderate): culated using the software INDICATE v. 1.2.0 (https://​ 11–20 m bufer strip, or 6–10 m bufer strip and fat or www.​syste​mecol​ogy.​de/​indic​ate). low slope; 4 (high): ≤ 10 m bufer strip, or ≤ 5 m bufer Application of other indicators of freshwater pollution strip and fat or low slope. To compare the performance and specifcity of Data analysis SPEAR­ in the identifcation of pesticide pollution, pesticides Data analysis was performed using R version 3.6.2. we additionally applied the Biological Monitoring Work- Because ­SPEAR has been developed for small and ing Party (BMWP) and the South African Scoring System pesticides medium streams with fowing water, we limited our anal- (SASS5) bioindicators for freshwater quality. In contrast ysis to samples from natural streams (no artifcial chan- to ­SPEAR , the BMWP and SASS5 indicators are pesticides nels) with an average width of < 20 m and an estimated based on a gradual instead of a binary classifcation sys- average fow velocity > 1 cm/s (based on the drift-body tem for the sensitivity of freshwater macroinvertebrates method). Two sites from a small river close to Kisii (PS42 (mainly to low oxygen levels) ranging from 1 to 10. Like and PS43, see Additional fle 2: Table S4) were excluded, SPEAR­ the indicators have been established for pesticides, because biological sampling took part during high water. freshwater families from a diferent region, but have been Tese samples were characterized by low numbers of also applied in sub-Saharan Africa beyond their countries taxa and individuals, as well as unusually high pesti- of origin [28, 39] cide concentrations as compared to the ­SPEARpesticides values, suggesting that sampling took place during a Ganatra et al. Environ Sci Eur (2021) 33:58 Page 6 of 15

run-of event. Terefore, we considered neither the toxicity (Fig. 2). Non- or marginally polluted sites with –3 SPEAR­ pesticides nor the pesticide data from this stream ­TUmax < ­10 were dominated by the ephemeropteran comparable to the other sites. Altogether, our analysis families Baetidae and Heptageniidae, the zygopteran included 16 sites from 13 diferent streams. Tree of the family Coenagrionidae, the heteropteran families Ger- streams were sampled twice, with an average distance of ridae and Veliidae, and by the coleopteran family Gyr- ca. 2 km between both sites. Macroinvertebrate samples inidae. Tese families occurred in at least half of the and toxic units from sites of the same stream were con- low-polluted streams and also showed the highest pro- siderably more similar than those from diferent streams; portions on the mean number of individuals collected in to avoid pseudo-replication, we, therefore, aggregated a low-polluted stream (relative abundances). Te same data from diferent sites of the same stream using the families occurred also in at least half of the highly pol- –3 mean values. luted streams with TU­ max ≥ ­10 , but additionally the Te original and the revised ­SPEARpesticides index, as ephemeropteran family Caenidae, the anisopteran fam- well as the associated estimated ­TUmax were compared to ily Libellulidae, the heteropteran families Naucoridae the ­TUmax from pesticide measurements using one-way and Nepidae, and snails of the family Planorbidae were linear regression; ­TUmax was log­ 10-transformed prior to observed in the majority of these streams. As the most the analyses. To assess the potential disturbance of the prominent changes in the community composition, the sampling sites by additional stressors other than pesti- mean relative abundance of Baetidae decreased from 56% cides, we investigated the distribution of the measured (range: 7–94%) in lowly polluted streams to 10% (range: ­TUmax and physicochemical water parameter values 0.2–38%) in highly polluted streams (n 5 lowly and 8 2 = using violin plots; the data were compared to thresholds highly polluted sites, χ = 8.29, p = 0.004 using a bino- for a good ecological status from the literature. To assess mial generalized linear model). At the same time, Planor- 2 the performance and the specifcity of ­SPEARpesticides for bidae increased from 0 to 9% (range: 0–36%, χ = 7.95, pesticides, we analyzed responses of the SPEAR­ pesticides p = 0.005) in highly polluted streams. Terefore, changes index and of the additional bioindicators and descriptors in the community composition with pesticide pollu- to the physicochemical parameters using one-way linear tion were driven by the response of frequently occurring regression. Additionally, relations of the ­SPEARpesticides taxa, suggesting that ­SPEARpesticides may perform consist- index and those physicochemical parameters that had ently across random samples from diferent streams. Te a (marginally) signifcant efect on the SPEAR­ pesticides, results are consistent with Reiber et al. [42] who identi- BMWP or SASS5 bioindicator were summarized using fed mayfies and snails among those macroinvertebrate a principal component analysis (PCA). Finally, we per- taxa that most strongly decrease or increase with pesti- formed pairwise correlations among the bioindicators cide pollution in European streams, respectively. and among those environmental variables that signif- cantly afected these bioindicators to identify confound- Adaptation of SPEARpesticides to East African streams ing factors. Most of the 35 macroinvertebrate families sampled were automatically linked to existing taxa in the trait database Results provided in the software INDICATE (see Additional Pesticide pollution and efects on the macroinvertebrate fle 2: Table S5). Only three families were not existing community in the database, because they do not occur in Central We observed considerable pesticide toxicity to fresh- European streams; these families were manually linked water macroinvertebrates in our 16 study sites from 13 to higher taxa available in the database: Pisuliidae was streams, quantifed as the maximum toxic unit derived linked to Trichoptera Gen. sp., Oligoneuridae was linked from chemical water analyses (TU­ max, see methods). to Ephemeroptera Gen. sp., and Chlorocyphidae was ­TUmax exceeded the threshold range for environmen- linked to Zygoptera Gen. sp. Culicidae was automati- tal efects of 10­ –4 to 10­ –3 [36] in nine and eight streams, cally linked to Diptera Gen. sp., but traits are missing for respectively. Te results indicate that pesticide pollution this taxon (due to a lack of data, because dipteran species was a relevant environmental stressor in the study area. are too heterogeneous to be aggregated), and they were, Pesticide toxicity was driven by the insecticides diazinon therefore, automatically excluded from the calculation of –2.36 –1.81 (7 sites, TU = ­10 –10 ), pirimiphos-methyl (3 sites, SPEAR­ pesticides. TU ­10–1.86–10–1.21), bendiocarb (1 site, TU ­10–2.54) Tough developed for European streams, the exist- = –3.47 = and imidacloprid (1 site, TU = ­10 ). ing ­SPEARpesticides index v. 2019.11 from Knillmann et al. Overall, we identifed 35 macroinvertebrate fami- [37] a showed reasonable correlation with the meas- lies in the streams. Te proportion of taxonomic ured pesticide toxicity ­(TUmax) from chemical water groups changed markedly with increasing pesticide analysis when applied to the Kenyan macroinvertebrate Ganatra et al. Environ Sci Eur (2021) 33:58 Page 7 of 15

Fig. 2 The Freshwater macroinvertebrate community composition changes in streams with high pesticide pollution. The stacked bars show the –3 mean proportions of macroinvertebrate families on the overall number of individuals observed in lowly polluted (maximum toxic unit ­TUmax < ­10 ) and in highly polluted streams ­(TU ­10–3). The number of individuals ranged from 32 to 154 (mean: 94) in the 5 lowly polluted streams and from max ≥ 37 to 462 (mean: 152) in the 8 highly polluted streams. Families that were recorded in at least 50% of the streams are shown in in bold. Taxa shown

in blue have been classifed as being at risk in the adapted ­SPEARpesticides indicator for East African streams, taxa shown in red have been classifed not at risk

2 and imidacloprid on freshwater macroinvertebrates col- communities(R = 0.43, Fig. 3a). Te correlation of the SPEAR­ pesticides index with toxic pressure increased when lected in the study region of Western Kenya revealed that we re-classifed two taxa based on additional informa- the tolerance of local coenagrionid species is comparable tion for Kenyan taxa. First, the damselfy family of Coe- to those of other taxa such as Notonectidae, Chironomi- nagrionidae has been classifed at risk in the European dae and Dytiscidae that have been classifed as insensitive SPEAR­ pesticides index v. 2019–11, based on a relatively high [9]. Accordingly, we re-classifed Coenagrionidae as being insensitive (s value 0.4); consequently, the taxon physiological insecticide sensitivity (s value = − 0.24). =− Tis value is only slightly above the threshold for sensitive moved from the category “at risk” to “not at risk”. Second, in SPEAR­ v. 2019.11, the heteropteran taxa (s ≥ − 0.36). Te classifcation was based on toxicity pesticides data for the European species Eschnura elegans. How- family Corixidae was considered as non-exposed due to ever, acute toxicity tests with the insecticides diazinon a surrounding layer of air (physical lung) that potentially Ganatra et al. Environ Sci Eur (2021) 33:58 Page 8 of 15

Fig. 3 SPEARpesticides indicates pesticide pollution in tropical streams of Western Kenya. Pesticide pollution was quantifed as the maximum toxic unit ­(TU , restricted to ­10–4) out of 30 pesticides measured in water samples collected during the rainy season but not during peak max ≥ exposure. Freshwater macroinvertebrates were identifed to the family level. (a) Application of the European SPEAR­ pesticides index v. 2019.11 from Knillmann et al. [37] to the Kenyan samples, R2 0.43, F 8.15, df 1, res. df 11, p 0.016, intercept 0.62, slope 0.37. (b) Application of the ======− ­SPEAR index after adaption to East African streams. R2 0.53, F 12.35, df 1, res. df 11, p 0.005, intercept 0.02, slope 0.43 pesticides ======− =−

protect individuals from pesticide exposure in the water. ratio of estimated vs. measured ­TUmax ranged from 0.2 to However, in acute toxicity tests we observed high insec- 19.6, i. e. the predicted toxicity deviated from the meas- ticide sensitivity of Corixidae from the study region [9]. ured toxicity by < 1.3 orders of magnitude in all investi- Tus, we re-classifed Corixidae as being exposed, so gated streams. that the taxon moved from the category “not at risk” to We related the East African SPEAR­ pesticides index “at risk”. Results of toxicity tests with additional 13 taxa to background pesticide toxicity measured in water collected in the study region from Becker et al. [9] were from grab samples collected during the rainy season, in accordance with the existing classifcation in the but outside run-of events. In contrast, in Europe the SPEAR­ pesticides trait data base. SPEAR­ pesticides index has been related to the short-term With these changes, we were able to adapt the peak toxicity measured during run-of events follow- SPEAR­ pesticides index to East African streams so that it ing heavy rainfall [20, 21, 37]. In temperate streams explained the observed ­TUmax considerably better than of Europe and North America, peak pesticide toxicity 2 the European SPEAR­ pesticides index v. 2019.11 (R = 0.53, is approximately four times as high as the background Fig. 3b). Te East African ­SPEARpesticides index decreased toxicity during the season of main pesticide exposure with increasing measured pesticide toxicity as follows: in early summer [15, 43–46]. Assuming that the same =− ∗ − ratio of peak vs. background toxicity may hold also in SPEARpesticides 0.43 log10(backgroundTUmax) 0.02 tropical streams, we applied a correction factor of 4 to (4) relate the East African ­SPEARpesticides index to an esti- After solving Eq. 4 for TU­ max, the East African mated peak exposure during run-of: SPEAR­ pesticides index may be used to predict background + SPEARpesticides 0.02 toxicity in streams from the observed macroinvertebrate −0.43 EstimatedpeakTUmax = 4 ∗ 10 (6) community composition: + SPEARpesticides 0.02 Te response of the East African ­SPEARpesticides index = −0.43 BackgroundTUmax 10 (5) to the estimated peak exposure in Kenya could then be compared to the response of the latest European Te mean TU­ predicted from Eq. 5 and the mean max SPEAR­ pesticides index v. 2019.11 to peak exposure in observed background ­TU coincided well (difer- max Germany [37]. Te East African ­SPEARpesticides index ence by a factor of < 2.5 or of ≤ 0.4 orders of magnitude produced considerably higher numbers as compared to (= ­log10-transformed ­TUmax) across the whole range the European index, particularly when pesticide toxic- of observed toxicity, Fig. 4a). Across all data points, the ity was low (Fig. 4b). Te more pronounced increase in the East African ­SPEARpesticides index with decreasing Ganatra et al. Environ Sci Eur (2021) 33:58 Page 9 of 15

Fig. 4 Performance of the East African ­SPEARpesticides index. (a) Background pesticide toxicity (maximum toxic unit, ­TUmax) predicted with the East African ­SPEAR index correlates with the measured background toxicity in Kenyan water samples. R2 0.54, F 12.97, df 1, res. df 11, pesticides = = = = p 0.004, intercept 0.41, slope 0.80. ­TU was restricted to ­10–4, because lower toxicity could not be measured due to detection limits for = =− = max ≥ pesticides. ­TUmax has been ­log10-transformed, so that diferences are presented in orders of magnitude. (b) Relation of the East African SPEAR­ pesticides index with the estimated peak pesticide toxicity during run-of in Kenyan streams, as compared to the relation of the European SPEAR­ pesticides index v. 219.11 with measured peak toxicity in German streams (data from [37]. Peak exposure in Kenyan streams was estimated by multiplying the measured background toxicity with a correction factor of 4 (see main text for justifcation)

pesticide exposure indicates diferences in the com- water parameters with values recommended for good position of European and East African macroinverte- water quality from the literature revealed that pesticide brate communities and in their response to pesticides. pollution was indeed one of the dominant stressors at Terefore, the use of a separate conversion scheme the sampling sites (Fig. 5). Additionally, in many streams from ­SPEARpesticides to TU­ max in afrotropical streams, as we observed very high levels of phosphate and turbidity, established with Eqs. 5 and 6, is justifed. as well as low levels of carbonate hardness and dissolved To assess the reproducibility and robustness of the East oxygen, which may have contributed in shaping the mac- African ­SPEARpesticides index, we extended our data set roinvertebrate community. to all sites sampled in the study area of Western Kenya Next, we assessed the sensitivity of the adapted (see [9]). Te 48 sites from 40 water bodies included SPEAR­ pesticidess index for East Africa to such additional habitats for which ­SPEARpesticides was not designed such stressors. Te ­SPEARpesticides index decreased signif- as artifcial irrigation channels, rice felds, reservoirs, cantly not only with increasing pesticide toxicity (meas- oxbow lakes and large rivers. We still observed a rela- ured ­TUmax and ­TUsum) and run-of potential, but also tion of the East African ­SPEARpesticides index with ­TUmax with carbonate hardness, conductivity and turbidity that was very similar to those in Fig. 3b, but the non- (Additional fle 1: Table S1). However, pesticide toxicity 2 explained variation increased considerably (R = 0.13, was signifcantly correlated with each of these confound- F = 5.55, df = 1, res. df = 38, p = 0.024, intercept = 0.13, ing factors (Additional fle 1: Table S2). All these factors slope = − 0.31). were associated with a dominant environmental gradient identifed by the frst principal component in a principal component analysis (PCA). Tis gradient explained 61% Specifcity of the East African SPEARpesticides index to of the total variation and likely refects a range of difer- pesticides ent stream types (Fig. 6). On one end, lowly polluted, fast- To assess the potential impact of confounding factors and fowing mountain streams showing high SPEAR­ pesticides the specifcity in the response of the adapted East Afri- values (streams nr. 27, 28, 31 and 32) were running in can ­SPEARpesticides index to pesticide pollution, we frst comparably steep valleys through intensely cultivated investigated the potential infuence of additional phys- tea plantations but were protected from run-of by wide icochemical parameters on the macroinvertebrate com- bufer strips so that run-of potential and pesticide tox- munity. A comparison of the observed physicochemical icity was low. On the other end, lowland streams were Ganatra et al. Environ Sci Eur (2021) 33:58 Page 10 of 15

Fig. 6 Principal component analysis of the East African ­SPEARpesticides index and those environmental variables that (marginally) signifcantly correlated with any of the investigated bioindicators. The frst principal component (horizontal axis) explains 60.7% of the overall variation and was considered to represent a gradient of stream types. The second principal component (vertical axis) additionally explains 14.7% of the overall variation and was considered to represent a gradient of high water level. Each point represents a stream (identifed by its number, see Additional fle 2: Table S4), colors

refer to the average ­SPEARpesticides value obtained from each stream

slowly fowing and showed higher run-of potential due to surrounding agriculture with only small bufer strips. Tese streams showed high pesticide toxicity, as well as sediment (turbidity) and phosphate pollution that were associated with low ­SPEARpesticides values and high con- ductivity, respectively. Tough all these confounding stressors increased simultaneously with the frst principal component (con- sidered as stream type), they could be disentangled along the second principal component (Fig. 6): On one end, Fig. 5 Distribution of physicochemical parameter values at the we observed slow fowing streams with clear water but study sites. The violin plot shows the kernel probability density of high conductivity and carbonate hardness. Streams on the data points across the parameter values. White points indicate the other end showed high fow velocity, turbidity and the median, black boxes the interquartile range, and black lines phosphate concentrations, but low conductivity and car- 1.5 the interquartile range. Light-grey boxes indicate the range of × values considered typical or recommend for streams with good water bonate hardness. We consider the second principal com- quality ponent to represent a gradient of raised water level at the day of sampling. Raised water levels may have increased fow velocity, as well as turbidity and phosphate levels due to erosion, and decreased conductivity and carbon- ate levels due to dilution with rainwater. In contrast, neither run-of potential, nor pesticide toxicity ­(TUmax) Ganatra et al. Environ Sci Eur (2021) 33:58 Page 11 of 15

or ­SPEARpesticides responded to this potential gradient of modifcations of SPEAR­ pesticides and the results from the water levels. First, run-of requires not only high water present case study. but also nearby agricultural felds. Second, raised water does not mean fooding in this context, which might Pesticide pollution in Western Kenya have been associated with pesticide pollution from run- Pesticides are an important stressor to freshwater mac- of (fooded sites were excluded from the analysis, see roinvertebrates in small and medium streams of west- methods). Te SPEAR­ pesticides index was thus most closely ern Kenya. Our results confrm earlier conclusions from associated with pesticide toxicity and run-of poten- various freshwater habitats of the same study area Kandie tial, indicating indeed a high specifcity of the adapted et al. [29], in that study, the chemical freshwater pollution SPEAR­ pesticides indicator for pesticide pollution in West- was assessed and the ecological risk ­(TUsum) identifed ern Kenyan streams. Each of the higher principal compo- was highest for macroinvertebrates due to insecticide nents explained only ≤ 8% of the total variance, and they exposure. Pesticide toxicity observed in streams of the were thus not further analyzed. present study was comparable to those in European land- Finally, we compared the performance of ­SPEARpesticides scapes characterized by intensifed agriculture [48]. In in indicating pesticide pollution to those of other bioindi- three out of 13 streams, we observed toxic units exceed- cators and commonly applied descriptors of the macroin- ing the threshold of 10­ –2 that is considered safe accord- vertebrate community. Pesticide toxicity most strongly ing to the frst tier of governmental risk assessment in correlated with the East African ­SPEARpesticides index the European Union [49]. Te results illustrate a need 2 (R = 0.53), followed by the European SPEAR­ pesticides for the monitoring and regulation of pesticide applica- index (R2 0.43) and the average score per taxon (aspt) tion to reduce pesticide exposure in freshwater. In Kenya, = 2 of the BMWP indicator (R = 0.33, Additional fle 1: plant protection products are sold at relatively low prices Table S1). Correlation with the aspt of the SASS5 indica- without the need for a certifcate of competence from 2 tor was lower (R = 0.19) and not signifcant. In contrast retailers, making them widely available to small farmers to ­SPEARpesticides that was only related to pesticides and who are then not informed of the necessary precautions their associated stressors (see above), the BMWP and needed to comply to the proposed environmentally safe SASS5 indicators were additionally related to phosphate use. Tis includes products containing active substances pollution and temperature but did not signifcantly cor- that have been banned in many high income countries, respond to run-of potential. Te EPT index (cumulative such as the non-selective insecticide diazinon that can be proportion of Ephemeroptera, Plecoptera and Trichop- applied on a wide range of crops (https://​ec.​europa.​eu/​ tera) showed a signifcant but low negative response to food/​plant/​pesti​cides/​eu-​pesti​cides-​datab​ase/) and was 2 pesticide pollution (R = 0.10), turbidity and carbonate driving toxicity in most of our sampling sites. hardness, but increased with fow velocity. Te Shannon index for species diversity increased with pesticide pollu- Adaptation of SPEARpesticides to East African streams tion, but the response was only marginally signifcant and Efects on the macroinvertebrate community composi- may relate to the generally concurrent increase in species tion were successfully quantifed even with the non-mod- richness and pesticide pollution from the spring to more ifed ­SPEARpesticides indicator v. 2019.11 developed for downstream sections [47]. Europe. ­SPEARpesticides values decreased approximately Te diferent bioindicators correlated with each other. log-linearly with increasing pesticide toxicity when the –4 –1.5 Te East African ­SPEARpesticides index increased with the maximum toxic unit (TU­ max) ranged from ­10 to ­10 . aspt of the BWMP and SASS5 indicator and with the EPT At lower ­TUmax values ­SPEARpesticides leveled of, con- –4 –3 index, but decreased with species diversity (Additional frming a threshold of ­10 to ­10 ­TUmax below which fle 1: Table S3). BMWP and SASS5 were clearly corre- no efects on freshwater macroinvertebrates have been lated with each other, but not with the EPT index and observed [36]. species diversity. Te EPT index decreased with increas- Reclassifcation of the families Coenagrionidae and ing species diversity. Corixidae based on toxicity tests with Kenyan species [9] improved the correlation of ­SPEARpesticides with ­TUmax. Discussion We suggest determining whether such a reclassifca- We adapted the ­SPEARpesticides bioindicator for the quan- tion may increase the link between toxicity and inverte- tifcation of pesticide exposure to afrotropical conditions brate community composition also in European streams. and demonstrated its use in thirteen Kenyan streams. A SPEAR­ pesticides values from both the European and the detailed step-by-step guidance for the application of the adapted East African index were considerably higher in East African ­SPEARpesticides index is provided in Addi- Kenya than those observed in Central European streams tional fle 1. In the following, we discuss the performed [37], particularly when pesticide toxicity was low. We Ganatra et al. Environ Sci Eur (2021) 33:58 Page 12 of 15

speculate that in East African streams, the natural pro- and dilution, the SPEAR­ pesticides index, pesticide pollu- portion of non-vulnerable taxa may be lower. Indeed, tion and run-of potential did not, because raised water macroinvertebrate communities in most Central Euro- levels alone do not increase run-of without nearby pean streams are dominated by amphipod crustaceans agricultural felds. Tus, our results show that indeed of the genus Gammarus sp., and by mayfies of the Bae- pesticide toxicity and not confounding factors is driving tis rhodani / vernus group [50]. Both groups are consid- the ­SPEARpesticides index. ered non-vulnerable to pesticides and often contribute In contrast to SPEAR­ pesticides, other bioindicators for the to more than 50% of individuals. In contrast, amphipod assessment of freshwater quality have not been designed crustaceans were missing in Kenya, confrming previous to specifcally indicate efects of pesticides [38, 40]. As observations e. g. from Elias et al. [51]. Additionally, we expected, the BMWP and SASS5 scoring system, there- classifed mayfy families observed in Kenya as vulnerable fore, responded to a broader range of stressors including (confrmed by a strong decrease with increasing pesticide phosphate and sediment pollution that may be associ- toxicity, and consistent with most other, less abundant ated with oxygen depletion. Similarly, the EPT index and European mayfies). Hence, ­SPEARpesticides values in East the Shannon index for species diversity most strongly Africa are associated to higher levels of pesticide toxic- responded to stressors other than pesticides. ity than similar SPEAR­ pesticides values in Europe, and we Our application of ­SPEARpesticides is not the frst attempt adapted the link between the East African SPEAR­ pesticides to establish or apply bioindicators for the assessment of index and ­TUmax accordingly. freshwater quality in East Africa [28, 52, 53]. Tese case Using the East African ­SPEARpesticides index, the meas- studies illustrate the growing interest in the use of bioin- ured pesticide toxicity ­(TUmax) could be predicted with dicators for freshwater monitoring but did not explicitly a precision of 1.3 orders of magnitude in each of the consider efects of a specifc stressor such as pesticide streams. Tis variation covered not only uncertainties in pollution. Apart from our study, the only application the ­SPEARpesticides approach, but also in the measurement of ­SPEARpesticides in sub-Saharan Africa we are aware of of pesticide concentrations and in their conversion to has been described in Malherbe et al. [22]. Te authors toxic units. Considering the generally high levels of varia- applied a previous version of the European SPEAR­ pesticides bility and uncertainty in ecotoxicology, as refected by an index and the Australian SPEAR­ pesticides index [23] to assessment factor of 100 (2 orders of magnitude) in frst macroinvertebrate samples from the Crocodile River tier European risk assessment [49], results from chemical and the Harts River in South Africa. Te sampling sites analyses and from the application of ­SPEARpesticides coin- were located upstream, adjacent to and downstream of cided reasonably well. two large irrigation schemes. Te SPEAR­ pesticides index decreased with increasing estimated pesticide toxicity Specifcity of SPEAR­ for efects of pesticides 2 pesticides (R = 0.26) but the correlation was not signifcant. Te overall variability in values of the East African It should be noted that Malherbe et al. [22] used a SPEAR­ pesticides index was explained to 53% by the meas- very coarse toxicity estimation based on the sampling ured pesticide toxicity, and to 45% by the estimated site location, assuming that pesticide pollution increases run-of potential based on catchment slope and width from upstream to downstream of the irrigation system. of bufer strips. Te estimated run-of potential was also Toxicity estimation thus did not consider pesticide input closely associated with pesticide toxicity and thus pro- from outside the irrigation scheme, whereas our data vides a fast and simple method for the identifcation of show that subsistence farming may considerably con- potential sampling sites of interest, and for the verifca- tribute to pesticide pollution. Terefore, the poor per- tion of calculated ­SPEARpesticides values: sites where the formance of ­SPEARpesticides in Malherbe et al. [22] may be estimated pesticide toxicity based on ­SPEARpesticides does partly related to uncertainties in the assessment of pes- not ft to the estimated run-of potential may be heavily ticide exposure. Additionally, the authors sampled mac- afected by additional stressors and should be investi- roinvertebrates from large rivers, whereas ­SPEARpesticides gated further. has been developed for small to medium streams. When Additional stressors that afected the East African we tested the East African ­SPEARpesticides index with a SPEAR­ pesticides index included carbonate hardness, con- larger data set of unsuitable habitats, our conversion ductivity and turbidity. All these confounding factors scheme from SPEAR­ pesticides values to pesticide toxicity increased with pesticide toxicity along a gradient of dif- turned out to be principally robust despite the small sam- ferent stream types but could be disentangled along a ple size used for development. However, the unexplained second, independent gradient of increasing water lev- variance considerably increased, illustrating the impor- els. While carbonate hardness, conductivity and turbid- tance to consider that the applicability of SPEAR­ pesticides ity varied with water levels presumably due to erosion Ganatra et al. Environ Sci Eur (2021) 33:58 Page 13 of 15

is limited to small and medium streams with fowing Funding This work was supported by the DFG (Deutsche Forschungsgemeinschaft), water and no heavy streambed degradation [21]. grant number LI 1708/4-1, BR 2931/3-1, HO 3330/12-1 and the German Helm- holtz long-range strategic research funding. We also gratefully acknowledge Conclusions the fnancial support for this research by icipe’s core donors, Foreign, Com- monwealth & Development Ofce (FCDO) of the UK Government; Swedish As shown, high impact of pesticides on freshwater International Development Cooperation Agency (Sida); the Swiss Agency for organisms is not limited to regions with intensifed Development and Cooperation (SDC); Federal Democratic Republic of Ethio- commercial agriculture. Widespread pesticide pollu- pia; and the Kenyan Government. tion in Western Kenyan streams and the associated Availability of data and materials decline in vulnerable macroinvertebrates indicate an All data generated or analysed during this study are included in this published ecological risk also in areas dominated by subsistence article [and its additional fles]. farming. Potential negative efects on species diversity and on ecosystem services such as leaf-litter degrada- Declarations tion and biological pathogen control illustrate the need Ethics approval and consent to participate to improve the risk management of pesticides also in Not applicable. developing countries. Monitoring is essential in this Consent for publication respect to identify hot spots of pesticide pollution for Not applicable. the targeted development of mitigation measures, and to evaluate the efectiveness of actions that have been Competing interests The authors declare no competing interests. taken. We adapted the ­SPEARpesticides bioindicator for the quantifcation of pesticide exposure in streams of Author details 1 East Africa. Tis tool provides a cost-efcient alterna- International Centre of Insect Physiology and Ecology (Icipe), Human Health Theme, Nairobi 00100, Kenya. 2 Biological Sciences, Egerton University, tive to the complex sampling and analysis of chemicals Egerton 20115, Kenya. 3 Department Evolutionary Ecology and Environmental and thus may facilitate large scale monitoring with lim- Toxicology, Institute of Ecology, Evolution and Diversity, Faculty Biological 4 ited resources in developing countries. Sciences, Goethe University Frankfurt, 60438 Frankfurt, Germany. Department of Biological Sciences, Moi University, P.O. Box 3900‑30100, Eldoret, Kenya. 5 Department of Efect-Directed Analysis, Helmholtz Centre for Environmental Research (UFZ), Permoserstrasse 15, 04318 Leipzig, Germany. 6 Department Abbreviations of System‑Ecotoxicology, Helmholtz Centre for Environmental Research GmbH BMWP: Biological Monitoring Working Party index; SASS5: South African Scor- – UFZ, 04318 Leipzig, Germany. 7 Institute for Environmental Research (Biology ing System 5; EPT: Ephemeroptera, Plecoptera & Trichoptera index; SPEAR pesti- V), RWTH Aachen University, 52074 Aachen, Germany. cides: Species at Risk index for pesticides; TU: Toxic Unit. Received: 19 February 2021 Accepted: 16 April 2021 Supplementary Information The online version contains supplementary material available at https://​doi.​ org/​10.​1186/​s12302-​021-​00497-9. References Additional fle 1. User guide and supplementary tables. 1. Sharma A, Kumar V, Shahzad B, Tanveer M, Sidhu GPS, Handa N, Kohli SK, Yadav P, Bali AS, Parihar RD, Dar OI, Singh K, Jasrotia S, Bakshi P, Additional fle 2. Tables with raw data. Ramakrishnan M, Kumar S, Bhardwaj R, Thukral AK (2019) Worldwide pesticide usage and its impacts on ecosystem. SN Applied Sciences 1(11):1446. https://​doi.​org/​10.​1007/​s42452-​019-​1485-1 Acknowledgements 2. Beketov MA, Keford BJ, Schaefer RB, Liess M (2013) Pesticides reduce We would like to thank the technical staf at icipe-TOC specifcally Paul Ouma regional biodiversity of stream invertebrates. Proc Natl Acad Sci USA and William Owigo for their help during data collection. We would like to 110(27):11039–11043. https://​doi.​org/​10.​1073/​pnas.​13056​18110 thank the developers of the free software statistical program R. All maps in 3. Geiger F, Bengtsson J, Berendse F, Weisser WW, Emmerson M, Morales were created using ArcGIS® software by Esri. ArcGIS® and ArcMap™ are the MB, Ceryngier P, Liira J, Tscharntke T, Winqvist C, Eggers S, Bommarco R, intellectual property of Esri and are used herein under license. Copyright © Pärt T, Bretagnolle V, Plantegenest M, Clement LW, Dennis C, Palmer C, Esri. All rights reserved. For more information about Esri® software, please Oñate JJ, Guerrero I, Hawro V, Aavik T, Thies C, Flohre A, Hänke S, Fischer C, visit http://​www.​esri.​com. Goedhart PW, Inchausti P (2010) Persistent negative efects of pesticides on biodiversity and biological control potential on European farmland. Authors’ contributions Basic Appl Ecol 11(2):97–105. https://​doi.​org/​10.​1016/j.​baae.​2009.​12.​001 ML conceived the approach; AG, FK, UF, FM, BT, WB, ML, HH & JMB designed 4. Sánchez-Bayo F, Wyckhuys KAG (2019) Worldwide decline of the entomo- the research; AG, FK and JMB conducted the research; JMB and AG analysed fauna: a review of its drivers. Biol Cons 232:8–27. https://​doi.​org/​10.​1016/j.​ and interpreted the data; JMB and AG drafted the initial version; AG, FK, UF, biocon.​2019.​01.​020 FM, BT, WB, ML & JMB contributed to the fnal version—all approved the fnal 5. Rundlöf M, Andersson GKS, Bommarco R, Fries I, Hederström V, Herberts- version of the publication. AG: Akbar A. Ganatra; FK: Faith Jebiwot Kandie; UF: son L, Jonsson O, Klatt BK, Pedersen TR, Yourstone J, Smith HG (2015) Ulrike Fillinger; FM: Francis McOdimba; BT: Baldwyn Torto; WB: Werner Brack; Seed coating with a neonicotinoid insecticide negatively afects wild ML: Matthias Liess; HH: Henner Hollert; JMB: Jeremias Becker. All authors read bees. Nature 521(7550):77–80. https://​doi.​org/​10.​1038/​natur​e14420 and approved the fnal manuscript. 6. Schaefer RB, Van den Brink PJ, Liess M (2011b). Impacts of pesticides on freshwater ecosystems. In: Ecological impacts of toxic chemicals. F. Sánchez-Bayo, P. J. Van den Brink and R. M. Mann. Bentham Science Publishers Ltd.: 111–137. Ganatra et al. Environ Sci Eur (2021) 33:58 Page 14 of 15

7. Roubos CR, Rodriguez-Saona C, Isaacs R (2014) Mitigating the efects of 27. Beketov MA, Foit K, Schaefer RB, Schriever CA, Sacchi A, Capri E, Biggs insecticides on arthropod biological control at feld and landscape scales. J, Wells C, Liess M (2009) SPEAR indicates pesticide efects in streams - Biol Control 75:28–38. https://​doi.​org/​10.​1016/j.​bioco​ntrol.​2014.​01.​006 Comparative use of species- and family-level biomonitoring data. Environ 8. Talebi K, Kavousi A, Sabahi Q (2008) Impacts of pesticides on arthropod Pollut 157(6):1841–1848. https://​doi.​org/​10.​1016/j.​envpol.​2009.​01.​021 biological control agents. Pest Technol 2(2):87–97 28. Ochieng H, Odong R, Okot-Okumu J (2020) Comparison of temperate 9. Becker JM, Ganatra AA, Kandie F, Mühlbauer L, Ahlheim J, Brack W, Torto and tropical versions of Biological Monitoring Working Party (BMWP) B, Agola EL, McOdimba F, Hollert H, Fillinger U, Liess M (2020) Pesticide index for assessing water quality of River Aturukuku in Eastern Uganda. pollution in freshwater paves the way for schistosomiasis transmission. Global Ecol Conserv 23:e01183. https://​doi.​org/​10.​1016/j.​gecco.​2020.​ Sci Rep 10(1):3650. https://​doi.​org/​10.​1038/​s41598-​020-​60654-7 e01183 10. Hernández AF, Parrón T, Tsatsakis AM, Requena M, Alarcón R, López-Guar- 29. Kandie FJ, Krauss M, Beckers L-M, Massei R, Fillinger U, Becker J, Liess nido O (2013) Toxic efects of pesticide mixtures at a molecular level: their M, Torto B, Brack W (2020) Occurrence and risk assessment of organic relevance to human health. Toxicology 307:136–145 micropollutants in freshwater systems within the Lake Victoria South 11. Münze R, Orlinskiy P, Gunold R, Paschke A, Kaske O, Beketov MA, Hundt Basin Kenya. Sci Total Environ 714:136748. https://​doi.​org/​10.​1016/j.​scito​ M, Bauer C, Schüürmann G, Möder M, Liess M (2015) Pesticide impact on tenv.​2020.​136748 aquatic invertebrates identifed with Chemcatcher® passive samplers and 30. Wambugu PW, Muthamia ZK (2009) Country Report on the State of the SPEARpesticides index. Sci Total Environ 537:69–80. https://​doi.​org/​ Plant Genetic Resources for Food and Agriculture. Nairobi, Kenya, Kenya 10.​1016/j.​scito​tenv.​2015.​07.​012 Agricultural Research Institute, National Genebank of Kenya 12. Morrissey CA, Mineau P, Devries JH, Sanchez-Bayo F, Liess M, Caval- 31. Pluskal T, Castillo S, Villar-Briones A, Orešič M (2010) MZmine 2: modular laro MC, Liber K (2015) Neonicotinoid contamination of global surface framework for processing, visualizing, and analyzing mass spectrometry- waters and associated risk to aquatic invertebrates: a review. Environ Int based molecular profle data. BMC Bioinformatics 11(1):395 74:291–303. https://​doi.​org/​10.​1016/j.​envint.​2014.​10.​024 32. Sprague JB (1970) Measurement of pollutant toxicity to fsh. II. Utilizing 13. Van Dijk T, Van Staalduinen M, Van der Sluijs J (2013) Macro-invertebrate and applying bioassay results. Water Res 4(1):3–32. https://​doi.​org/​10.​ decline in surface water polluted with imidacloprid. PLoS ONE 8:5. 1016/​0043-​1354(70)​90018-7 https://​doi.​org/​10.​1371/​journ​al.​pone.​00623​74 33. PPDB - The Pesticide Properties Database (2019) Agriculture & Environ- 14. Von Der Ohe PC, Liess M (2004) Relative sensitivity distribution of aquatic ment Research Unit AERU, University of Hertfordshire. http://​sitem.​herts.​ invertebrates to organic and metal compounds. Environ Toxicol Chem ac.​uk/​aeru/​iupac/​index.​htm. Accessed 26 Mar 2019. 23(1):150–156. https://​doi.​org/​10.​1897/​02-​577 34. Brock TCM, Van Wijngaarden RPA (2012) Acute toxicity tests with Daphnia 15. Liess M, Schulz R, Liess MHD, Rother B, Kreuzig R (1999) Determination of magna, Americamysis bahia, Chironomus riparius and Gammarus pulex insecticide contamination in agricultural headwater streams. Water Res and implications of new EU requirements for the aquatic efect assess- 33(1):239–247. https://​doi.​org/​10.​1016/​s0043-​1354(98)​00174-2 ment of insecticides. Environ Sci Pollut Res 19(8):3610–3618. https://​doi.​ 16. Beketov MA, Schäfer RB, Marwitz A, Paschke A, Liess M (2008) Long-term org/​10.​1007/​s11356-​012-​0930-0 stream invertebrate community alterations induced by the insecticide 35. ECOTOXicology Knowledgebase (2019) U.S. Environmental Protection thiacloprid: efect concentrations and recovery dynamics. Sci Total Envi- Agency (USEPA), Ofce of Pesticide Programs, Environmental Fate and ron 405(1–3):96–108. https://​doi.​org/​10.​1016/j.​scito​tenv.​2008.​07.​001 Efects Division, Ecological Efects Branch. http://​cfpub.​epa.​gov/​ecotox/​ 17. Cold A, Forbes VE (2004) Consequences of a short pulse of pesticide help.​cfm?​help_​id ​DATAS​TEWAR​D&​help_​type ​defn​e&​help_​back 1. exposure for survival and reproduction of Gammarus pulex. Aquat Toxicol Accessed 2019–03–26.= = = 67(3):287–299. https://​doi.​org/​10.​1016/j.​aquat​ox.​2004.​01.​015 36. Schaefer RB, von der Ohe PC, Rasmussen J, Keford BJ, Beketov MA, Schulz 18. Liess M, Schulz R (1999) Linking insecticide contamination and popula- R, Liess M (2012) Thresholds for the efects of pesticides on invertebrate tion response in an agricultural stream. Environ Toxicol Chem 18(9):1948– communities and leaf breakdown in stream ecosystems. Environ Sci 1955. https://​doi.​org/​10.​1002/​etc.​56201​80913 Technol 46(9):5134–5142. https://​doi.​org/​10.​1021/​es203​9882 19. Vrana B, Allan IJ, Greenwood R, Mills GA, Dominiak E, Svensson K, Knuts- 37. Knillmann S, Orlinskiy P, Kaske O, Foit K, Liess M (2018) Indication of pes- son J, Morrison G (2005) Passive sampling techniques for monitoring ticide efects and recolonization in streams. Sci Total Environ 630:1619– pollutants in water. TrAC Trends Anal Chem 24(10):845–868 1627. https://​doi.​org/​10.​1016/j.​scito​tenv.​2018.​02.​056 20. Liess M, Schafer RB, Schriever CA (2008) The footprint of pesticide stress 38. Dickens CW, Graham PM (2002) The South African Scoring System in communities - species traits reveal community efects of toxicants. Sci (SASS) Version 5 Rapid Bioassessment Method for Rivers. Afr J Aquat Sci Total Environ 406(3):484–490. https://​doi.​org/​10.​1016/j.​scito​tenv.​2008.​05.​ 27(1):1–10. https://​doi.​org/​10.​2989/​16085​914.​2002.​96265​69 054 39. Bere T, Nyamupingidza BB (2014) Use of biological monitoring tools 21. Liess M, von der Ohe PC (2005) Analyzing efects of pesticides on inver- beyond their country of origin: a case study of the South African Scoring tebrate communities in streams. Environ Toxicol Chem 24(4):954–965. System Version 5 (SASS5). Hydrobiologia 722(1):223–232. https://​doi.​org/​ https://​doi.​org/​10.​1897/​03-​652.1 10.​1007/​s10750-​013-​1702-7 22. Malherbe W, van Vuren JHJ, Wepener V (2018) The Application of a Mac- 40. Paisley MF, Trigg DJ, Walley WJ (2014) Revision of the Biological Monitor- roinvertebrate Indicator in Afrotropical Regions for Pesticide Pollution. J ing Working Party (BMWP) score system: derivation of present-only and Toxicol 2018:2581930–2581930. https://​doi.​org/​10.​1155/​2018/​25819​30 abundance-related scores from feld data. River Res Appl 30(7):887–904. 23. Schaefer RB, Keford BJ, Metzeling L, Liess M, Burgert S, Marchant R, https://​doi.​org/​10.​1002/​rra.​2686 Pettigrove V, Goonan P, Nugegoda D (2011) A trait database of stream 41. Schriever CA, von der Ohe PC, Liess M (2007) Estimating pesticide runof invertebrates for the ecological risk assessment of single and combined in small streams. Chemosphere 68(11):2161–2171. https://​doi.​org/​10.​ efects of salinity and pesticides in South-East Australia. Sci Total Environ 1016/j.​chemo​sphere.​2007.​01.​086 409(11):2055–2063. https://​doi.​org/​10.​1016/j.​scito​tenv.​2011.​01.​053 42. Reiber L, Knillmann S, Foit K, Liess M (2020) Species occurrence relates to 24. Hunt L, Bonetto C, Marrochi N, Scalise A, Fanelli S, Liess M, Lydy MJ, pesticide gradient in streams. Sci Total Environ 735:138807. https://​doi.​ Chiu MC, Resh VH (2017) Species at Risk (SPEAR) index indicates efects org/​10.​1016/j.​scito​tenv.​2020.​138807 of insecticides on stream invertebrate communities in soy production 43. Kreuger J (1998) Pesticides in stream water within an agricultural catch- regions of the Argentine Pampas. Sci Total Environ 580:699–709. https://​ ment in southern Sweden, 1990–1996. Sci Total Environ 216(3):227–251. doi.​org/​10.​1016/j.​scito​tenv.​2016.​12.​016 https://​doi.​org/​10.​1016/​S0048-​9697(98)​00155-7 25. Menezes S, Baird DJ, Soares AMVM (2010) Beyond taxonomy: a review 44. Münze R, Hannemann C, Orlinskiy P, Gunold R, Paschke A, Foit K, Becker of macroinvertebrate trait-based community descriptors as tools for JM, Kaske O, Paulsson E, Peterson M, Jernstedt H, Kreuger J, Schüürmann freshwater biomonitoring. J Appl Ecol 47(4):711–719. https://​doi.​org/​10.​ G, Liess M (2017) Pesticides from wastewater treatment plant efuents 1111/j.​1365-​2664.​2010.​01819.x afect invertebrate communities. Sci Total Environ 599–600:387–399. 26. Wang Z, Kwok KWH, Lui GCS, Zhou G-J, Lee J-S, Lam MHW, Leung KMY https://​doi.​org/​10.​1016/j.​scito​tenv.​2017.​03.​008 (2014) The diference between temperate and tropical saltwater species’ 45. Schäfer RB, Paschke A, Vrana B, Mueller R, Liess M (2008) Performance of acute sensitivity to chemicals is relatively small. Chemosphere 105:31–43. the Chemcatcher® passive sampler when used to monitor 10 polar and https://​doi.​org/​10.​1016/j.​chemo​sphere.​2013.​10.​066 semi-polar pesticides in 16 Central European streams, and comparison Ganatra et al. Environ Sci Eur (2021) 33:58 Page 15 of 15

with two other sampling methods. Water Res 42(10):2707–2717. https://​ 51. Elias J, Ijumba J, Mgaya Y, Mamboya F (2014) Study on freshwater doi.​org/​10.​1016/j.​watres.​2008.​01.​023 macroinvertebrates of some tanzanian rivers as a basis for developing 46. Williams RJ, Brooke DN, Matthiessen P, Mills M, Turnbull A, Harrison RM biomonitoring index for assessing pollution in tropical African Regions. J (1995) Pesticide transport to surface waters within an agricultural catch- Ecosyst 2014:1–8. https://​doi.​org/​10.​1155/​2014/​985389 ment. Water Environ J 9(1):72–81. https://​doi.​org/​10.​1111/j.​1747-​6593.​ 52. Masese FO, Raburu PO, Muchiri M (2009) A preliminary benthic mac- 1995.​tb009​28.x roinvertebrate index of biotic integrity (B-IBI) for monitoring the Moiben 47. Minshall GW, Petersen RC, Nimz CF (1985) Species richness in streams River, Lake Victoria Basin Kenya. Afr J Aqu Sci 34(1):1–14. https://​doi.​org/​ of diferent size from the same drainage basin. Am Nat 125(1):16–38. 10.​2989/​AJAS.​2009.​34.1.​1.​726 https://​doi.​org/​10.​1086/​284326 53. Shimba MJ, Jonah FE (2016) Macroinvertebrates as bioindicators of water 48. Becker JM, Russo R, Shahid N, Liess M (2020) Drivers of pesticide resist- quality in the Mkondoa River, Tanzania, in an agricultural area. Afr J Aquat ance in freshwater amphipods. Sci Total Environ 735:139264. https://​doi.​ Sci 41(4):453–461. https://​doi.​org/​10.​2989/​16085​914.​2016.​12305​36 org/​10.​1016/j.​scito​tenv.​2020.​139264 49. EFSA PPR (2013) Guidance on tiered risk assessment for plant protection products for aquatic organisms in edge-of-feld surface waters. EFSA J Publisher’s Note 11(7):3290. https://​doi.​org/​10.​2903/j.​efsa.​2013.​3290 Springer Nature remains neutral with regard to jurisdictional claims in pub- 50. Becker JM, Liess M (2017) Species Diversity Hinders Adaptation to Toxi- lished maps and institutional afliations. cants. Environ Sci Technol 51(17):10195–10202. https://​doi.​org/​10.​1021/​ acs.​est.​7b024​40