Landscape simplification and altitude affect biodiversity, herbivory and Andean potato yield

Authors

  • Katja Poveda,

    Corresponding author
    1. Agroecology, Department of Crop Sciences, Georg-August University, Griesbachstraße 6, 37077 Göttingen, Germany, and Department of Entomology, Cornell University, 4142 Comstock Hall, Ithaca, NY 14853. USA
      Correspondence author: Department of Entomology, Cornell University, 4142 Comstock Hall, Ithaca, New York 14853. USA. E-mail: kap235@cornell.edu
    Search for more papers by this author
    • †These authors contributed equally to this work.

  • Eliana Martínez,

    1. Biology Department, Universidad Nacional de Colombia, Bogotá, Colombia
    Search for more papers by this author
    • †These authors contributed equally to this work.

  • Monica F. Kersch-Becker,

    1. Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, NY, USA
    Search for more papers by this author
  • Maria A. Bonilla,

    1. Biology Department, Universidad Nacional de Colombia, Bogotá, Colombia
    Search for more papers by this author
  • Teja Tscharntke

    1. Agroecology, Department of Crop Sciences, Georg-August University, Griesbachstraße 6, 37077 Göttingen, Germany, and Department of Entomology, Cornell University, 4142 Comstock Hall, Ithaca, NY 14853. USA
    Search for more papers by this author

Correspondence author: Department of Entomology, Cornell University, 4142 Comstock Hall, Ithaca, New York 14853. USA. E-mail: kap235@cornell.edu

Summary

1. The simplification of agricultural landscapes through the increase in cropped area has caused the loss of habitats for many species that fulfil important ecosystem services such as pest control and production. Evidence for detrimental effects on ecosystem services is scarce, particularly in tropical regions.

2. We studied the effect of the percentage of cropped land in the landscape and altitude in tropical agro-ecosystems in relation to crop pest regulation and yield. In the Colombian Andes, we established potato Solanum tuberosum plots along gradients of altitude and increasing proportion of cropped area to assess the effects on herbivores, their natural enemies, potato production and overall biodiversity.

3. Increasing altitude and percentage cropped land reduced the richness and abundance of herbivores and their natural enemies, except for the specialist Guatemalan potato moth Tecia solanivora, which showed the opposite response.

4. Potato yield was negatively affected by the presence of the Guatemalan potato moth, which increased in density as the percentage of cropped land and altitude increased. Other herbivores and natural enemies did not affect yield.

5.Synthesis and applications. Tropical landscapes at lower altitude or with smaller areas of cropped land suffered less from the presence of the potato moth, which had a negative effect on yield. Our results suggest that conservation of natural habitats like the endangered Andean ecosystems would benefit farmers through ecosystem services such as reduced pest damage, higher yield and increased functional biodiversity.

Introduction

Agro-ecosystems have faced great changes in the last decades owing to increased conversion of natural habitat to cropped lands and through the intensification of agriculture, both of which are drivers of biodiversity loss (Krebs et al. 1999; Sala et al. 2000; Tilman et al. 2001). Changes may occur at different spatial scales including increased agrochemical inputs at a local scale and a reduction in landscape complexity at a wider scale as more land is cultivated (Tscharntke et al. 2005). As agro-ecosystems rely heavily on ecosystem services such as pollination and natural pest control, landscape simplification and the loss of functionally important biodiversity are of major global concern (United Nations 2005). For example, natural pest control has been positively associated with the diversity of natural enemies (Wilby & Thomas 2002; Gurr, Wratten & Luna 2003; Snyder & Ives 2003). However, the role of biodiversity in maintaining natural pest control, decreasing herbivore pressure and increasing productivity is less well understood (but see Cardinale et al. 2003; Ostman, Ekbom & Bengtsson 2003; Finke & Denno 2004). At a landscape scale, structurally complex landscapes with small areas of cropped land interspersed between natural habitat have been found to increase natural enemy populations and reduce pest pressure on crops in 74% and 45% of cases, respectively (based on a review by Bianchi, Booij & Tscharntke 2006). Based on the resource concentration hypothesis, this pattern would be expected, given that a higher percentage of cropped land could be of benefit to herbivores feeding on crops (Root 1973). However, none of the studies reviewed the effect on crop yield (Bianchi, Booij & Tscharntke 2006), and we are not aware of any more recent investigations on landscape complexity and the effects on crop productivity. Although there is evidence that landscape structure can influence crop damage (Thies & Tscharntke 1999), it remains unclear whether the effect on natural enemies and pests cascades down to crop yield. Apart from the study by Parsa, Canto & Rosenheim (2011), we are not aware of any other study carried out in the highly diverse tropical ecosystems; most studies examining the relationship between landscape complexity and diversity have been performed in less diverse temperate regions. Although temperate regions undergone landscape transformation over long periods of time, tropical ecosystems have faced this change increasingly in the last decades (Sodhi, Brook & Bradshaw 2007). The relationship between landscape complexity and ecosystem functions such as natural pest control and productivity in those highly biodiverse ecosystems could provide some insights into the functional importance of biodiversity in changing tropical environments.

Biodiversity is affected by natural factors as well as by human activities. Environmental gradients such as altitude are also known to affect biodiversity and may interact with anthropogenic forces that shape ecosystem function. Higher altitudes are associated with a decline in species richness in many groups of animals, including insects (Lawton, Macgarvin & Heads 1987; Hodkinson 2005). However, in the Andes, the evidence for a reduction in pest pressure with altitude is not based on field data, but on observations of the preference shown by farmers to grow crops at higher elevations to avoid the negative effects of herbivores and pathogens (Montaldo 1984). The tendency to convert high-elevation natural areas to cropped land has led to widespread clearance of endangered ecosystems such as the paramos in Colombia, where 60% of the total area has been converted to agricultural use such as cattle farming and potato fields (Monasterio, Smith & Molinillo 2006; Rangel 2006). In the higher Colombian Andes, potato Solanum tuberosum, L. is the most important crop. Approximately 170 000 ha distributed in 250 municipalities between 2000 and 3500 m are planted with potato, employing directly more than 110 000 families (Osorio, Espitia & Luque 2001; Lopez 1980).

In this study, we focused on the effect of percentage cropped land and altitude on functional arthropod diversity, including herbivores and their natural enemies, in Andean potato fields. We tested the direct and indirect effects of landscape variables, herbivores and natural enemies on potato yield. We proposed three alternative hypotheses for the mechanism behind this pattern: (i) Altitude and the percentage of cropped land in a 1-km-radius landscape area are directly related to potato yield; (ii) Following the natural enemies’ hypothesis, both increased cropped land and altitude negatively affect the diversity and abundance of natural enemies, releasing herbivores from top-down predation pressures and decreasing yield; (ii) Following the resource concentration hypothesis, increased cropped land positively affects herbivore abundance and potato herbivory, which has a negative impact on potato tuber yield.

Materials and methods

Study Area and Planting

We planted a total of 17 potato S. tuberosum cv Pastusa Suprema plots in the Cundinamarca Department of Colombia (Fig. S1). We established six 20 × 12 m plots between July and December 2006 and 11 plots between March and July 2007. The plots were situated along a gradient from landscapes with intensive land use (94% agricultural area) to more complex landscapes with a high percentage of natural habitats (37% agricultural area). Plots were also established along an altitudinal gradient between 2584 and 3227 m above sea level, unrelated to the landscape complexity gradient (r = 0·34, = 0·17, n = 17). Distance between plots ranged from 1·2 to 66·1 km with a mean distance (±SE) of 30·3 ± 1·6 km. Each plot consisted of 20 furrows, 12 m in length and separated by 1 m. Seed tubers were planted every 35 cm. Potato seed tubers were planted between June and July 2006 and in March 2007. Plants were fertilized twice. During establishment, they were fertilized with N : P : K 12 : 34 : 12 (Ecofertil S.A., Bogota, Colombia). At the time of ridging (also known as hilling) 8 weeks after establishment, they were fertilized with N : P : K 10 : 10 : 10 (Ecofertil S.A., Bogota, Colombia). Seedlings of potato plants are normally attacked by flea beetles (Coleoptera: Chrysomelidae) that can cause complete defoliation and death of the plants in the first weeks after emergence (Poveda K. & E. Martinez personal observation). To control for this initial stages of foliar herbivory but without compromising the survivorship of the organisms associated with the potato crops, we sprayed the plants weekly with a commercial garlic–pepper extract (Agrisan, Cundinamarca, Colombia; 120 mL per 20 L) for the first 8 weeks and before the plant started producing tubers. We favoured garlic–pepper extract over any commercial insecticide, as its main way of action is insect repellence and not toxicity. Our previous work has shown that spraying garlic–pepper extract does not affect the predator community in comparison with unsprayed plots, while insecticide treatments reduce predator and overall arthropod abundance (Gomez-Jimenez & Poveda 2009). Insects were first sampled 4 weeks after the last garlic–pepper spray. We controlled late blight Phytophthora infestans (Mont.) de Bary through biweekly spraying of Dithane M-45® (Dow Agrosciences de Colombia, Bogota, Colombia; active compound: Mancozeb 80%), alternated with Fitoraz WP 76® [Bayer CropScience S.A. Bogota, Colombia; active compound: Propineb (70%) and Cymoxanil (6%)].

Landscape Data

At the landscape scale, we mapped different land use types (potato, pasture and other crops) and different natural habitats (exotic forest, native forest and hedgerows) through field visits and inspection of aerial photographs and official topographic maps [Instituto Geográfico Agustin Codazzi (IGAC) scale 1 : 10 000 for aerial photographs and 1 : 5000 for cartographic material] using the program gis arcview 3.2 (ESRI 1999). We used the percentage cropped land (crops and heavily grazed pastures dominated by the tropical grass species Pennisetum clandestinum Hoechst Ex Chiov.) in a circular area of 1 km radius surrounding each study site as a measure of landscape simplicity (see Tscharntke et al. 2005). This measure was correlated with the Shannon habitat-type diversity (r = −0·87 < 0·0001 n = 17: calculated from the six types of habitats measured) and marginally correlated with the percentage area of potato fields (r = 0·46, P = 0·058, n = 17). We chose to measure percentage cropped land over the percentage area of potato fields because potatoes are usually rotated with pastures and potato pests that finish their development in the soil can easily move from recent pastures to potato fields. To correct for the effect of the relief on the area estimation, we topographically corrected all aerial photographs through the program erdas imagine (Leica Geosystems 2001) using 20 control points. There was a gradient in land use intensity from 37% to 94% cropped area with a mean ± SE of 74 ± 17·6% across the 17 plots (for more details see Appendix S1).

Soil Factors

We determined local soil chemistry in each field by taking five soil cores of 10 cm depth during the potato harvesting period (December 2006 and August 2007). Samples were analysed for percentage nitrogen and organic carbon by the soil laboratory of the Agronomy Faculty of the Universidad Nacional de Colombia, Bogota.

Arthropod Sampling

We caught epigeic fauna using four pitfall traps (7 cm in diameter and 10 cm height) arranged at the four corners of a 8-m square. We opened the traps for two 1-week sampling periods during potato flowering and 2 weeks before harvesting. Sampling dates were chosen based on the previously reported increased attraction to herbivores of the potato plants at the moment of tuberization (Niño 2004), which starts with the flowering period. The trapping fluid was an alcohol: water dilution (2 : 1) with some added detergent to break surface water tension. Arthropods were separated by order and identified to family.

We sampled flying and leaf-dwelling arthropods between two central furrows on the days that pitfall traps were opened and emptied. We took sweep-net samples the day pitfall traps were opened and the day those traps were collected. We summed the four pitfall trap and four sweep-net subsamples to obtain the total number of arthropods for each of the sampling dates.

According to previous reports that emphasize the importance of assigning arthropods to different functional groups (i.e. Clough, Kruess & Tscharntke 2007b; Tscharntke et al. 2007), we divided the arthropod families found into predators, parasitoids, herbivores, saprophages and omnivores (based on what is reported by Carrejo & Gonzalez 1992; Triplehorn & Johnson 2005; Wolf 2006). Organisms that could not clearly be assigned to a trophic group were included in the analysis of total number of families and total abundance, but not the analysis of the different functional groups. All analyses were conducted at a family level because of the lack of identification keys to identify insects to a more precise level.

The Guatemalan potato moth Tecia solanivora Povolny (Lepidoptera: Gelechiidae) is one of the main potato insect pests in Latin America (OEPP & EPPO 2005), so we estimated its presence in 2007 by setting up pheromone traps in the centre of each plot. These traps are effective up to a distance of 30 m and capture only males; this underestimates the actual population size (Niño 2004) but is adequate for performing comparisons between plots. We left traps in the field for 7 days at the end of the cultivation period (August 2007), when the population should be at its highest peak. Data are reported as the number of individuals captured per day.

Potato Yield

To quantify potato yield, we harvested 20 plants from the centre of each plot, when plants began to senesce. We determined the number and weight of undamaged (marketable) and damaged (unmarketable) tubers per plant by visual inspection after cutting open each tuber. Damage was mainly caused by T. solanivora, larvae, although we found evidence of damage by two other pests, Premnotrypes vorax Hustache (Coleoptera: Curculionidae) and Naupactus sp (Coleoptera: Curculionidae). To estimate pest pressure, we calculated the percentage of tubers damaged by dividing the number of damaged tubers by the total number of tubers per plant.

Data Analysis

Adequacy of sampling effort was validated by calculating species accumulation curves and the Incidence-based Coverage Estimator (ICE) using estimates, Version 8.2.0 (Colwell 2006) with 500 randomizations. The degree of saturation at a family level was indicated by the percentage of observed arthropod families relative to the estimated family richness and was 83·1% for all samples. These results suggest that sample size and sampling effort were sufficient.

To evaluate the effects of cropped area and altitude on arthropod abundance and richness at a family level, as well as potato yield, we used generalized linear models with a stepwise backward simplification to comply with the principle of parsimony (Crawley 2003). The analysis was performed with the program statistica (Statsoft 2003), and all variables not significant at < 0·05 were excluded from the model. The percentage of cropped land was used as a measure of landscape simplicity, and altitude was included as a measure of the altitudinal gradient. Year of sampling was used as a random factor, and soil nitrogen and soil organic matter were included as covariates in the analysis. To adjust to normal distribution of residuals and heteroscedasticity, we ln-transformed the variables herbivore and parasitoid family richness, total number of arthropods, herbivores, predators, parasitoids and T. solanivora. We plotted the standardized residuals and fitted to a normal curve to determine adjustment to normal distribution. The inspection of plots showing standardized residuals against fitted values showed us which data were more homogeneous in variance (Crawley 2003). The percentage of potatoes that were damaged was arcsin square-root-transformed. Model assumptions about spatially independent and identically distributed residuals were checked by Mantel tests and diagnostic plots in R (The R project for statistical computing, version 2.10, http://www.r-project.org/). No spatial autocorrelation was found for any models reported in Table 1 (−0·31 < r < 0·18; > 0·14).

Table 1.   Results of generalized linear models with a stepwise backward simplification of the percentage cropped land in a 1 km radius around the plot, altitude and year of study in relation to the family richness and abundance of arthropod fauna associated with potatoes Solanum tuberosum and potato production
Response variablePercentage cropped landAltitude of plotYear of studyModel fit
  1. Statistics for each significant variable (ns, not significant) and the fit of the whole model for each variable are presented. Arthropod fauna was divided into three major functional groups: herbivores, predators and parasitoids. Total arthropod fauna included saprophages and omnivores.

Arthropod family richnessF1,14 = 8·05
= 0·013
F1,14 = 24·34
< 0·001
nsR2 = 0·77
F2,14 = 23·98
< 0·001
Herbivore family richnessF1,15 = 8·83
= 0·009
nsnsR2 = 0·32
F1,15 = 8·83
= 0·009
Predator family richnessnsF1,14 = 10·67
= 0·006
F1,14 = 4·96
= 0·042
R2 = 0·48
F2,14 = 8·67
= 0·004
Parasitoid family richness (log)F1,14 = 9·6
= 0·008
F1,14 = 13·7
= 0·002
nsR2 = 0·67
F2,14 = 17·82
< 0·001
Arthropod abundance (log)nsF1,15 = 23·01
< 0·001
nsR2 = 0·58
F1,15 = 23·01
< 0·001
Herbivore abundance (log)F1,14 = 6·6
= 0·022
F1,14 = 7·86
= 0·014
nsR2 = 0·56
F2,14 = 11·1
= 0·001
Predator abundancensF1,15 = 9·41
= 0·008
nsR2 = 0·34
F1,15 = 9·41
= 0·008
Parasitoid abundance (log)nsF1,15 = 16·52
= 0·001
nsR2 = 0·49
F1,15 = 16·52
= 0·001
Pests abundance (log)F1,14 = 8·59
= 0·01
F1,14 = 7·95
= 0·013
nsR2 = 0·59
F2,14 = 12·68
< 0·001
% damaged tubers (arcsin sqrt)nsnsF1,15 = 5·36
= 0·035
R2 = 0·21
F1,15 = 5·37
= 0·03
Weight undamaged potatoes (log)nsnsF1,15 = 5·06
= 0·039
R2 = 0·20
F1,15 = 5·06
= 0·039

Given that we had data on the Guatemalan potato moth abundance for 2007 but not for 2006, we investigated the effect of potato moth abundance on percentage tubers damaged and on marketable tuber weight through Spearman’s correlations. We also analysed how altitude, percentage cropped area and previous plot use affected potato moth abundance through a generalized linear model (glm) with a stepwise backward simplification. These analyses were all done with the program statistica (Statsoft 2003).

Direct and Indirect Effects of Landscape and Arthropod Variables on Potato Yield

To investigate whether percentage cropped land or altitude affected final marketable potato yield and whether it was a direct or an indirect effect driven through herbivores or natural enemies, we used path analysis to test our hypotheses. In our path model, we included percentage cropped land, as a measure of landscape simplification, altitude, natural enemies (richness and abundance), herbivores (richness and abundance) excluding the potato moth, as well as T. solanivora abundance. Using our data on the abundance of the potato moth in 2007, we plotted the regression line between the percentage tubers damaged (arcsin square-root-transformed for a better fit) and T. solanivora abundance in 2007 (T. solanivora abundance = −9·58 (% tubers damaged) +12·39; R2 = 0·43, F1·9 = 6·89, = 0·027) from which we estimated our predicted T. solanivora abundance. We used percentage of tubers damaged as a proxy for T. solanivora abundance in 2006. Although the percentage of tubers damaged in 2006 was much lower than in 2007, the data points were still within the range of data used to calculate the regression between percentage of tubers damaged and T. solanivora abundance in 2007. To ensure that the damage on potato tubers reflected T. solanivora abundance alone and not the abundance of other pests in general, we correlated the abundance of potato pests with the percentage of tubers damaged and with potato yield. In both cases, we found no correlation between the variables (Spearman’s correlations: r = −0·29, = 0·25, n = 17: r = 0·12, = 0·63, n = 17). Although there is some information about the natural enemies of T. solanivora (López-Ávila & Espitia-Malagón 2000) including both parasitoids and predators, we only made identifications to the family level. In addition, our results show that predators and parasitoids are similarly affected by percentage cropped land and altitude. Therefore, we decided to pool data on predators and parasitoids together in a ‘natural enemy’ category, reducing the number of paths in our model and increasing our statistical power. Herbivore abundance, natural enemy abundance and richness as well as potato yield were ln-transformed prior to the path analysis to meet parametric assumptions. The path analysis was calculated in the program systat (Systat Software Inc, Chicago, IL, USA 2004).

Results

Effects on Arthropods

A total of 25 892 individuals from 16 orders and 113 families (See Appendix S2) were captured. Most of the arthropods were saprophages (64·1%), followed by herbivores (24·8%), predators (6·1%) and parasitoids (4·4%). We sampled a total of 9270 herbivores and their natural enemies, classified into 13 orders and 69 families.

Both altitude and percentage cropped land within a 1 km radius around the plot had a negative relationship with both arthropod and parasitoid family richness and nearly all functional groups (Table 1, Fig. 1). Only predators did not respond to changes in the percentage cropped land, while herbivore family richness did not vary along the altitudinal gradient (Table 1, Fig. 1).

Figure 1.

 Family richness of total arthropods (a & b), herbivores (c & d), predators (e & f) and parasitoids (g & h) collected on potato fields in the Colombian Andes in relation to the cropped land area in a circular area of 1 km radius around the plot (a, c, e & g) and the altitude at the site of each plot (b, d, f & h).

The abundance of all arthropod groups was negatively correlated with altitude, but only herbivores were negatively affected by an increasing percentage of cropped land (Table 1, Fig. 2).

Figure 2.

 Abundance of total arthropods (a & b), herbivores (c & d), predators (e & f) and parasitoids (g & h) collected in potato fields in the Colombian Andes in relation to the percentage of cropped land in a circular area of 1 km radius around the plot (a, c, e & g) and the altitude at the site of each plot (b, d, f & h).

Potato Yield

Percentage cropped land and altitude had no direct effect on potato production, but there was an effect of year on the percentage of potatoes damaged and the weight of marketable potatoes (Table 1). In 2007, the mean (±SE) percentage of potato tubers damaged (mainly by the Guatemalan potato moth) was 50% (±8%), whereas in 2006, it was around 16% (±11%). At the same time, the mean (±SE) marketable potato yield in 2006 was 1324 (±215) g compared to 337 (±159) g in 2007.

The damage caused to the potato tubers was mainly inflicted by the Guatemalan potato moth Tecia solanivora. Data on T. solanivora abundance for 2007 showed a strong relationship between the abundance of male moths in the region and the damage to tubers (r = 0·62; = 0·039, Fig. 3a) and a negative relation to marketable potato yield (ln-transformed; r = −0·70; = 0·015, Fig. 3b). Abundance of T. solanivora was positively affected by the percentage agricultural area (Fig. 4a) and by altitude (Fig. 4b) and was higher when the plot was previously cultivated with potatoes compared with plots previously put to a different use (Table 2, Fig. 4c). Altitude and percentage cropped land were not correlated with each other (r = 0·45, = 0·16).

Figure 3.

 Relationship between the abundance of Tecia solanivora and the percentage of tubers damaged (a) and the yield (weight of marketable potato tubers) (b) for 2007 data.

Figure 4.

 Mean number of Tecia solanivora moths captured per day in relation (a) to the percentage of cropped land in a circular area of 1 km radius around the plot, (b) to the altitude and (c) to the previous use of the plots (five plots previously planted with potatoes; six plots previously pastures).

Table 2.   Effects of the percentage of cropped land in a 1 km radius around the plot, altitude and previous potato cover on the abundance of the potato moth Tecia solanivora
Variable assessedPercentage cropped landAltitude of plotPrevious coverModel fit
  1. F, P and R2 values as well as the degrees of freedom for each response variables are reported after performing a generalized linear model.

Log (Tecia solanivora abundance +1)F1,7 = 23·3
= 0·002
F1,7 = 249·9
< 0·001
F1,7 = 243·8
< 0·001
R2 = 0·97
F3,7 = 135·3
< 0·001

The path analysis of how different variables affected potato yield showed that it was strongly reduced as the abundance of T. solanivora increased (Fig. 5). The fit of the model did not differ between the two models (richness and abundance) (RMSEArichness < 0·001, = 0·800; RMSEAabundance < 0·001, = 0·722). Altitude and per cent cropped land did not affect potato yield directly, but both variables indirectly decreased production by increasing the incidence of potato moth. Higher densities of T. solanivora were recorded at higher altitudes and in landscapes with higher percentages of cropped land.

Figure 5.

 Path diagram for the model of the effect of percentage of cropped land in a circular area of 1 km radius around the plot, altitude, natural enemies, herbivores and T. solanivora abundance on potato yield. Path diagram including a) species richness for natural enemy families and herbivores and b) the abundance of natural enemies and herbivores. Solid lines denote significant effects, and dashed lines denote non-significant effects. Width of each line is proportional to the strength of the relationship. *< 0·05, **< 0·01, ***< 0·001.

At higher altitudes, the abundance and species richness of natural enemies decreased, but they were not affected by the percentage of cropped land. Landscapes with a higher percentage of cropped land harboured lower abundance and lower species richness of herbivorous insects (excluding T. solanivora), but the abundance of these herbivores did not alter potato yield. Both path diagrams showed a positive relationship between the abundance and species richness of natural enemies and the incidence of T. solanivora, but natural enemies did not affect the abundance and richness of herbivores.

Discussion

Both the percentage of cropped land in the landscape and altitude directly influenced the abundance of herbivores and their natural enemies. Our results showed that a decline in diversity at a landscape scale can increase pest abundance and consequently can have a negative effect on crop production. Potato yield decreased with increased potato moth abundance. Tecia solanivora populations increased with percentage cropped land, suggesting that low-diversity, simplified landscapes encouraged the presence of the moth, while natural enemies did not reduce its abundance. This finding supports the resource concentration hypothesis over the natural enemies hypothesis as the most plausible mechanism for the variation found in T. solanivora abundance. These results are the first field data to support the commonly claimed notion that high-diversity crop habitats are more resilient to pest species outbreaks and this potentially can have an indirect positive influence on crop yield.

The long-term sustainability of ecosystems and the services they generate depend on the conservation of biodiversity at scales beyond the local scale (Bengtsson et al. 2003; Loreau, Mouquet & Holt 2003). Landscapes with a low incidence of cropped lands (with a large arthropod species pool) should support a more diverse and abundant natural enemy community thus maximizing natural pest control and crop productivity (Tscharntke et al. 2007). Our data show that as the percentage of cropped land in the landscape surrounding the plot increased, there was a negative effect on parasitoid richness. Landscape simplification (high percentage of cropped land and low habitat-type diversity) is a result of agricultural intensification causing loss of the natural habitats that provide resources for the arthropod fauna (Kruess 2003; Schmidt & Tscharntke 2005; Tscharntke et al. 2005). Parasitoid longevity and fecundity, for example, are dependent on the availability of food resources like pollen and nectar as well as their hosts in the environment (Baggen & Gurr 1998; Tylianakis, Didham & Wratten 2004; Wackers 2004). In complex habitats, these types of resources are normally found adjacent to the main crops in non-cropped vegetation where they support beneficial organisms (Freeman Long et al. 1998).

Altitude had a negative effect on the overall arthropod family richness. The linear reduction or hump-shaped distribution of arthropod species richness along an altitudinal gradient has been reported previously, and several hypotheses have been proposed to explain this phenomenon (reviewed by Hodkinson 2005). Lower plant diversity at higher altitudes can limit the abundance and species richness of herbivores and therefore also the abundance of natural enemies (Hodkinson 2005). Moreover, a series of abiotic factors such as lower temperatures and increased UV-B radiation can directly or indirectly affect insect richness and abundance (Coulson & Whittaker 1978; Roff 1980; McCloud & Berenbaum 1994; Bird & Hodkinson 1999).

According to the resource concentration hypothesis (Root 1973), we would expect that plots surrounded by simple landscapes (higher percentage cropped land) would have high abundance of herbivores. However, in temperate regions, it has been reported that simplified landscapes have a lower abundance of herbivores (Roschewitz et al. 2005; Thies, Roschewitz & Tscharntke 2005; Clough, Kruess & Tscharntke 2007a; Rand & Tscharntke 2007). Similarly, in our study, we found that a higher percentage of cropped land reduced the abundance and richness of herbivores. This may be due to high levels of disturbance in agro-ecosystems and increased use of insecticides.

The potato moth is a specialist potato pest whose larvae can only feed on potato tubers (OEPP & EPPO 2005), yet it had a different impact on yield in successive years. Damage caused to the potatoes by the potato moth was 16% in 2006 and 50% in 2007. In 2007, we found more moths in plots surrounded by a high percentage of cropped land, suggesting that landscape simplification had a positive effect on potato moth abundance. The results from the path analysis showed that the abundance of the potato moth is more strongly affected by bottom-up effects providing support for the resource concentration hypothesis, as opposed to top-down effects where natural enemies would be controlling the moth. Simpler landscapes contain a higher percentage of potato crops, which makes the landscape more attractive for specialized pests and provide more food resources for larger pest population. Moreover, the abundance of traditional, relatively open potato shelters, where potato farmers store the tubers, is higher in areas with more intensive potato production. Potato shelters offer optimal conditions for moth development (Keasar et al. 2005) and could be the main source of potato moth infestation in the field. Although our regression data suggest that the potato moths encounter fewer natural enemies in landscapes with higher percentage of cropped land, the path analysis illustrates that the natural enemies recorded in this study are not controlling T. solanivora abundance, thus providing no support for the natural enemy hypothesis.

The adaptation of T. solanivora to higher altitudes parallels the adaptation of its host plant, the potato, to high altitudes in the Andes (Lopez 1980). However, it has been shown that temperature can limit the development of T. solanivora under laboratory as well as under field conditions (Dangles et al. 2008). Along an altitudinal gradient from 2550 to 3650 m in Ecuador, lower temperatures at higher altitudes reduced the abundance of T. solanivora (Dangles et al. 2008). This effect was not observed in our study probably because our altitudinal gradient reached a maximum of 3227 m rather than the 3650 m recorded in Dangles et al. (2008). This suggests that the abundance of T. solanivora may peak at mid-elevations and that abiotic factors such as temperature could constrain its upper distribution.

Finally, when analysing the effects of altitude and percentage of cropped land on potato tuber damage and marketable yield, we did not detect a general effect of either factor on productivity, yet we found a clear difference between study years. The percentage of tubers damaged differed between the 2 years, indicating differences in abundance of the potato moth. In 2007, we found a strong correlation between the percentage of cropped land and the abundance of potato moths, as well as between the abundance of potato moths and the percentage of tubers damaged at harvest and the final marketable yield of the potato plants. Nevertheless, there was no significant direct effect of the percentage of cropped land on yield. Instead, our analysis showed a pest-mediated reduction in potato yield as the percentage of cropped land and altitude increased.

In conclusion, we found the landscape context and altitude to be important in shaping the arthropod community richness and abundance. Increased percentage of cropped land and altitude led to a decrease in arthropod species richness and abundance, and the effect was similar in all feeding guilds (except for T. solanivora), contrary to what would be expected by functional-group-specific responses. We provide the first field data showing how a decline in tropical landscape diversity owing to increased percentage of cropped land is related to an increased pest abundance, which in turn has an effect on crop production. Tecia solanivora is the main potato pest causing high tuber damage and reducing crop productivity. We suggest that the conservation of natural habitats like the endangered high-altitude Andean ecosystems is not only important to conserve biodiversity per se but also to conserve the ecosystem functions provided by natural biodiversity.

Acknowledgements

We thank A. Ramirez, C. Ñustez, E. Torrado, E.A. Pedraza, P.A. Diaz, J. Jácome, Y. Ospina, P. Gonzalez, D. Campos, H. Aguirre, O. Dix, J.M. Vargas and the 17 farmers who collaborated with the project for assistance with field or laboratory work and advice. We specially thank the ‘Laboratorio de Entomología’ and the ‘Grupo de Investigación en Papa’ of the Agronomy Faculty – Universidad Nacional de Colombia for logistical support. We thank S. Mc Art, D. Gabriel, S. Parsa, J. Cepeda and O. Dangles for their helpful comments on the manuscript. This research was funded by the German Research Foundation (DFG) through grants PO 1215_2·1 and PO 1215_3·1 to K. P.

Ancillary