Pesticides reduce tropical amphibian and reptile diversity in agricultural landscapes in Indonesia

Introduction

The continuing growth of the human population requires an increase in agricultural productivity to secure the rising demand for food, fibre and bioenergy. This demand has led to agricultural expansion, mainly in the tropics, which has led to subsequent biodiversity loss (Foley et al., 2011). Consequently, there is an ongoing debate about how agricultural production in the tropics can produce higher yields and yet also protect biodiversity (Wanger et al., 2020; Tamburini et al., 2020). Some argue that intensifying production on existing agricultural land will allow conservation of biodiverse areas (‘land sparing’), while others contend that conservation value of agricultural land can be improved to protect biodiversity and ecosystem services (‘land sharing’; Grass et al., 2019). In practice, the decision will typically depend on the specific location and species groups (Kremen, 2015). For example, amphibians and reptiles are threatened by land-use change, but some species appear to survive well under land-sharing scenarios (Wanger et al., 2010a).

Land-sharing will result in pesticide exposure for amphibians and reptiles with negative impacts known primarily from laboratory studies for herbicides such as atrazine and glyphosate and the insecticide endosulfan (Berger, Graef & Pfeffer, 2013; Grant, Woudneh & Ross, 2013; Hayes et al., 2010; Relyea, Schoeppner & Hoverman, 2005). These substances are mostly banned in the European Union (EU) but are still commonly used in tropical countries, where little is known about the effect of increasing pesticide application on amphibians and reptiles (Wanger, Rauf & Schwarze, 2010). A review investigated the effects of pesticides on amphibians and reptiles based on 1,336 studies and 23,152 comparisons (Ortiz-Santaliestra et al., 2017), and showed that species exposed in mesocosm studies were 1,429 times more sensitive than in field experiments when lethal effect were considered. However, information about pesticide toxicity for terrestrial amphibians and reptiles in general is limited (but see Brühl et al., 2013; Agostini et al., 2020) and, hence, also precludes extrapolating results from other terrestrial taxa such as birds and mammals (Ortiz-Santaliestra et al., 2017). Given the lack of knowledge from European and North American systems, it is not surprising that this pattern is exacerbated in a tropical context. To the best of our knowledge, long-term data and field experiments to predict effects of pesticide application on tropical amphibians and reptiles in agricultural landscapes are missing.

Addressing this data gap, we built on a unique opportunity in cocoa production landscapes in Indonesia, where pesticide use started at the onset of our study. We combine a correlative field survey across five habitat types (primary forest, secondary forest, cacao agroforests, cacao monocultures, and clear-cut grassy areas) to evaluate how pesticide use affects amphibians and reptiles on the plot and landscape scale in Sulawesi, Indonesia. Pesticide use on the landscape scale may prevent non-target organisms from avoiding pesticide exposure and will affect food sources. We also experimentally exposed amphibians and reptiles to realistic pesticide application patterns in the field. The combined correlative-experimental field approach and our two-taxon comparison allowed a realistic assessment of the effects of agricultural intensification on biodiversity and the implications for ecosystem-service provisioning. We tested the hypothesis that both, amphibian and reptile richness and abundance will be measurably affected by commonly applied insecticides and herbicides.

Methods

Results

The pesticide survey to determine long-term pesticide use revealed that the Indonesian farmers used herbicides and not insecticides on the 36 study plots for the correlative approach (note that this is not the case in the experimental study area where insecticides are regularly applied; Motzke et al., 2013). Specifically, they applied combinations of the herbicides glyphosate and paraquat (three owners used glyphosate, five owners used paraquat, six owners used both). Herbicide use across habitats in the study area increased by more than six-fold, from an average use of 0.3 litres per hectare (l/ha) to 2 l/ha of the same dosage, from 2007 to 2008 (see Fig. S1). Our first sampling in 2007 was able to serve as a temporal control, because (i) pesticide use in all habitats but clear cut grassy areas was close to zero; and (ii) locals reported that pesticide use was negligible before sampling started in the study area. Overall, we recorded 12 amphibian and 15 reptile species (Tables S2A and S2B).

The correlative approach showed that plot-level variables were the best predictors of amphibian richness and abundance in cocoa agroforestry landscapes (Table 1). Amphibian richness (intercept = −0.55 ± 0.48; freq = −0.38 ± 0.20 reported are mean coefficients and SE) and abundance decreased with an increasing herbicide application frequency (intercept = −0.13 ± 0.55; freq = −0.32 ± 0.16; Fig. S2B). However, these effects are not statistically robust, because the null model also ranked amongst the best models (Δelpd < 4; Sivula, Magnusson & Vehtari, 2020; Table 1). Abundance of the generalist species, Ingerophrynus celebensis and the common forest species, Hylarana celebensis, was not predicted by either plot- or landscape-level predictors (null model was the model with the lowest edpd _loo; Table 1). As a noteworthy observation, we found many I. celebensis and Fejervarya limnocharis (Fig. S2A) with deformed limbs in the intensively sprayed plots of our study area. By contrast, neither reptile species richness nor general or Eutropis multifasciatus abundance was predicted by plot- or landscape-level variables (Table 2; Fig. S2C).

The experimental application of pesticides in Indonesian vegetable plantations was detrimental to both amphibians and reptiles. Amphibian richness and abundance, and abundance of the I. celebensis decreased in all treatments compared with the control (i.e., Bayesian credibility intervals of the control and treatment estimates do not overlap; Fig. 1 and Fig. S3). I. celebensis is the same species that was identified as a pest control agent of invasive ants in the same study area (Wanger et al., 2010b). The herbicide and the combined treatments—but not the insecticide—led to reduced reptile species richness compared with the control. All pesticide treatments reduced total reptile abundance, and the abundance of the common reptile species Eutropis spp. compared with the control (Fig. 1 and Fig. S3).

Discussion

Combining 1-year field observations and a field experiment, we found a strong negative response of tropical amphibians and reptiles to pesticide use in tropical agricultural landscapes in Indonesia. Amphibians showed higher sensitivity to pesticide exposure than reptiles.

In this study, amphibians were exposed to the herbicides glyphosate and paraquat, and the insecticide cypermethrin. Most amphibian ecotoxicological data on glyphosate, the world’s most widely used herbicide, is from aquatic rather than terrestrial life stages (for a review see Wagner et al., 2013), including mixture effects with other pesticides (Annett, Habibi & Hontela, 2014; Lanctôt et al., 2014). Glyphosate can lead to altered developmental rates, physical malformations, and premature death (Lanctôt et al., 2014; Mann et al., 2009). Laboratoy studies have shown that paraquat is lethal to tadpoles, suppresses reproductive success, and is of high genotoxicity (Quassinti et al., 2009; Yin et al., 2008). Information on cypermethrin (insecticide) effects on tropical amphibians is scarce, but it acts synergistically with glyphosate further decreasing tadpole survival (Brodeur, Poliserpi & Sánchez, 2014).

Our results also showed that herbicide (i.e., a mix of glyphosate and paraquat) application decreased amphibian species richness at the plot scale (although it also led to decreased abundance). The richness decrease might have resulted from a removal of sensitive specialist species such as Limnonectes spp. from the community. Although not statistically robust, the negative effect of application frequency on amphibian abundance corresponded with reported glyphosate impacts on amphibians (Lanctôt et al., 2014; Mann et al., 2009). The lack of detectable 1-year effects on the I. celebensis could be explained by its high mobility and ability to move into the forest, thereby lowering exposure (Wanger et al., 2011). A lack of effects on the most common forest species (H. celebensis) might be due to limited exposure on the edges of cocoa plantations. In contrast, direct experimental exposure of amphibians to paraquat, cypermethrin, and the combination treatment negatively affected most species, including the common I. celebensis, under realistic field conditions. The difference of results between correlative and experimental field effect studies suggests that more such combined studies in particular with long term (>1 yr) correlative components are required to better understand the underlying reasons of such differences.

Glyphosate studies on reptiles are limited, being largely restricted to laboratory investigations (Latorre et al., 2013; Poletta et al., 2011), and we did not find any studies documenting the effect of paraquat on reptiles. Cypermethrin is known to cause genotoxicity, enzymatic and metabolic alterations, and growth delay in caimans (Poletta et al., 2011). As neither plot nor landscape variables showed a detectable effect on general reptile richness and abundance or E. multifasciatus abundance, perhaps because the concentrations were too low to affect this more resilient taxon with less permeable skin compared to amphibians (Weir et al., 2016). In contrast, in the experimental approach, paraquat and the combined paraquat/cypermethrin exposure reduced reptile abundance and richness, and eliminated common species. Less sensitive reptiles might move from areas with high to low pesticide load, again emphasizing the importance of landscape-level management. Cypermethrin also reduced reptile abundance and common species abundance. No effects on reptile richness from these chemicals could be the result of immigration of less sensitive species such as Eutropis rudis and E. cf. grandis into treated plots (Wanger et al., 2011), attracted by lower competition from absent more sensitive species.

Our results suggest that increasing pesticide use in our study region may decrease amphibian pest-control services. In previous work from the same study areas, we showed that I. celebensis can influence cocoa yield through a direct negative effect on invasive ants (Anoplolepis gracilipes) and an indirect positive effect on native ants (Wanger et al., 2010b). As invasive ants reduce cocoa yield by 34%, the common toads (I. celebensis) can act as an indirect cocoa pest control agents (Motzke et al., 2013; Wanger et al., 2010b). As commonly used pesticides eliminate these toads, but have both positive and negative effects on different ant species (Motzke et al., 2013), we speculate that farmers may lose amphibian pest-control services, and so paradoxically have more pests that they then may control with more pesticides (Fig. S5).

Conclusions

In conclusion, our results show the strong effects that pesticides can have on both groups—under real field conditions. A better understanding of pesticide effects on tropical amphibians and reptiles is needed, because extrapolating existing results from temperate to tropical regions led to highly inconsistent results, and could only be based on freshwater and arthropod species (Daam & VandenBrink, 2010; Kwok et al., 2007; Daam et al., 2019). Moreover, regulations for pesticide registration in the United States and the EU do not yet require data on amphibians and reptiles (e.g., EFSA et al., 2018). As tropical developing countries typically base their regulatory processes on those already in place in the United States or EU, pesticide-effect studies on amphibians and reptiles both in the temperate zone and in the tropics will help to protect these groups more effectively, in some of the most biodiverse regions on Earth.

Supplemental Information