Pest Control Services from Natural Habitat

Insect surveys
Syrphids showed a strong response to landscape complexity in all three years of the study, increasing significantly with natural habitat in the area surrounding the farm (Fig. 1, r2 = 0.21, p <<0.01 for 2006; r2 = 0.35, p <<0.01 for 2007; r2 = 0.37, p <<0.01 for 2008). The metrics and scales at which these predators were most strongly correlated with landscape were not consistent from year to year, however. In the first year, proportion of non-crop habitat (which can include residential, and industrial land-uses in additional to natural habitat) was more predictive than proportion of natural habitat alone. In the second and third years, proportion of natural habitat was most predictive. In the first and third years, syrphids were most tightly correlated with landscape at a scale of 1000 meters, while in the second year the correlation was strongest at a scale of 2500 meters. Age of planting was also a significant factor (p < 0.05 in all years), with syrphids increasing toward harvest, but date during the growing season was not. Natural enemy richness, measured as the number of different natural enemy taxa per site each week, also increased with natural habitat in 2007 and 2008 (Fig. 2, r2 = 0.28, p <<0.01 for 2007; r2 = 0.43, p <<0.01 for 2008), but this trend was not significant in 2006 (p = 0.12). Again, the relevant scales varied, 1000 meters being most predictive for natural enemies in 2008 and 2000 meters being most predictive in 2007. Plant age and date during the growing season were also significant factors in most years, as natural enemy communities grew more diversified over the time to harvest (p < 0.01 for 2006 and 2008; non-significant in 2007) and over the course of the summer (p < 0.05 for all years). Parasitism was not correlated with landscape in any of the years (Fig. 3). This is an interesting result, as other studies have shown increased parasitism in addition to increased enemy abundance and richness in landscapes with greater landscape complexity(4). This difference could be due to the fact that the complex end of the landscape gradient in this study contains far more natural habitat than most studies, and as specialist enemies, parasitoids may have a harder time finding their prey in too diverse of landscapes. Parasitism appears to peak in the mid range of the landscape gradient, between 20 and 40% natural habitat in the surrounding 3000 meters. While parasitoids are reliant on flowering plants as nectar sources, they do not benefit from alternate prey, and may therefore require less off-farm habitat than more generalist natural enemies. The majority of studies investigating the effect of landscape on so-called pest control stop short of actually measuring the effect on the pests themselves. An increase in natural enemy abundance, richness, or activity (such as parasitism rate) is often taken as an indicator of increased pest control, without actually documenting a decrease in pest abundance or damage(5). In this study, no significant effect of landscape complexity was found on aphids in 2006 or 2007, and in 2008 the trend found was the reverse of what was expected, aphids increasing slightly with proportion of natural habitat in 1500 meters of the farm (Fig. 4, r2 = 0.26, p <0.05). This analysis masks a temporal interaction with landscape, however, which is better illustrated by Figure 5. Breaking the landscape complexity down into categories reveals that aphid distributions behave more as expected toward the ends of the landscape than in the middle. Despite similar initial aphid densities, farms in “simple” landscapes (with less than 15% natural habitat in the surrounding 1500 meters) showed greater aphid increases for the latter half of the growing period than farms in “complex” landscapes (surrounded by more than 50% natural habitat) in all three years. As pest infestation becomes more problematic for farmers closer to harvest, this seems a relevant way to examine the data. The mid range of the landscape gradient (20-50% natural habitat in 1500 meters) is more confounding; aphid densities exceeded those at either end of the gradient in 2006 and 2008, while in 2007 aphid densities were lower until the final week before harvest, when there appeared to be a population explosion. Using these somewhat arbitrary though useful categories to break down the landscape gradient can be helpful in elucidating further trends. Examining syrphid abundance and natural enemy richness in these landscape categories week by week reveals a similar trend as that found for the aphids. That is, the difference between syrphid densities (Fig. 6) and natural enemy diversity (Fig. 7) at the ends of the gradient is fairly pronounced in all three years, but measures for these natural enemies in the mid range of the gradient do not necessarily fall neatly between the two extremes. Furthermore, as is seen with aphid densities, the difference between natural enemies at the two ends of the gradient changes over time. However, while for aphids the two ends of the gradient began with similar densities that grew faster at the simple (<15% natural habitat) end than the more complex (>50% natural habitat) end, the reverse appears to be true for natural enemies. Farms in complex landscapes had more syrphids and more diverse enemy communities in the first weeks of the growing period than those in simple landscapes in all three years, and the two generally converged over time, with the simple landscapes closing the gap and sometimes even surpassing those complex landscapes by harvest time. This apparent lag time in the appearance of natural enemies at the farms in simple landscapes may account for the greater aphid growth seen toward harvest at these sites. Aphid population growth is exponential in absence of predation, and if natural enemies do not arrive early on enough in the establishment phase, aphid populations could easily grow beyond the point where enemies can contain them.

The effect of landscape on the response of syrphids to the previous week’s densities of aphids is also an informative way to examine these data. It makes sense to compare syrphids in one week to aphids in the previous week, because the implicit hypothesis in studying pest control is that natural enemies arrive in response to pest densities, and therefore we would expect higher syrphid densities the week after seeing higher aphid densities. In 2008 and 2006, complex landscapes showed a much steeper response than simple landscapes; syrphids increased more for a given density of aphids in the previous week on farms in complex landscapes (Fig. 8; r2 = 0.47, p <<0.01 for 2008; r2 = 0.17, p <0.05, with an landscape by aphid interaction of p << 0.01 for 2006). The difference was even more pronounced in 2006, when farms in simple and mid-range landscapes show a completely flat response, with a similar number of syrphids regardless of the previous week’s aphid densities. In 2007, the syrphid response to aphids has a similar slope in simple landscapes as in complex landscapes, but the intercepts are offset such that there are nearly double the syrphids for a given density of aphids in complex landscapes as simple landscapes (r2 = 0.27, p <<0.01). The mid-range landscapes again defy generalizations across years, showing a steeper slope than even the complex landscapes in 2008 and a shallower one than both simple and complex landscapes in 2006 and 2007. While it becomes increasingly difficult to understand trends in the mid-range landscapes as the data are more closely examined, the contrast between the two extremes of the landscape gradient is fairly apparent. Farms in complex landscapes begin their growing period with greater natural enemy densities and diversity as compared to simple landscapes, this difference is magnified by a stronger response to existing aphid populations in complex landscapes, and may account for higher aphid peaks toward harvest in simple landscapes, despite potentially similar initial densities. However, two issues remain unresolved. The first is why the trend for aphid densities is only apparent at the ends of the landscape gradient, when syrphid densities and enemy richness both show fairly strong positive correlations with increasing landscape complexity. Aphid distributions are obviously governed by more factors than natural enemies alone, such as climate, wind patterns, and source habitat. It could be that these other mitigating factors overwhelm the effect of enemies on aphid populations in the mid-range landscapes. Identifying one possibility for an additional mitigating factor (source habitat) is the focus of the last section, on the physiological impacts of mustard. A second unresolved issue is that while slower natural enemy response may account for the near-harvest aphid peaks in simple landscapes, an observational study such as this one is inadequate to assign cause and effect. In order to truly determine whether natural enemies are better containing pests in complex landscapes as compared to simple landscapes, a manipulative experiment is needed. This is addressed in the following section, the field experiment assessing pest control function. Field experiment: pest control function
Pest control was assessed at the site level, subtracting the average aphid densities on the sentinel plants from the average aphid densities on the plants in the open cages, and dividing that by the aphid densities on the plants in the closed cages. Subtracting this value from 1 yields the proportional reduction in aphid densities due to pest control. Considering a relative measure such as pest reduction as opposed to final aphid densities controls for endogenous differences in aphid population growth rate at different sites. As aphid reproduction is strongly dependent on temperature, climatic differences between sites can result in vastly different growth rates. Comparing aphid population growth in the closed cages (free from predation) reveals this disparity, with sites ranging from a mere doubling of aphid densities over the 12 day study period to a ten-fold increase. These differences in reproductive rates were not related to landscape complexity.

Pest control provided by natural enemies in complex landscapes (>50% natural habitat) was nearly double that found in simple landscapes (<15% natural habitat), with pest suppression approaching (and at one site even exceeding) 100% in complex landscapes (Figure 9a, Single-Factor ANOVA, p << 0.01). Pest suppression in excess of 100% means that the plants in the open cages had lower aphid densities than even the sentinel plants; that is, enemies were able to reduce not only the experimental aphid population, but they eliminated any naturally settling aphids as well. The sentinels most likely did not receive comparable pest control services because there was no initial aphid population to attract enemies to these plants early in the experiment. While the effect of landscape complexity on pest control is significant, the variability is not consistent. As shown by the standard error bars in Figure 9a, the error or statistical noise in the data for simple landscapes is more than twice that for complex landscapes. A closer inspection of the data reveals that two of the simple sites exhibited much greater pest reduction than the other two. Interestingly, this difference corresponds to a difference in local farm diversity. Although not a part of the intended experimental design, the sites within each landscape category break down neatly into two local categories: those with hedgerows adjacent to the field and diverse plantings (locally complex), and those lacking hedgerows and employing large-scale monoculture broccoli (locally simple). Reanalyzing the data along these lines (Two-Factor ANOVA with local and landscape complexity, see Figure 9b) shows that there is a significant landscape effect (p << 0.01), a significant local effect (p << 0.01), and a significant interaction between the landscape and local scales of complexity (p = 0.01). Specifically (according to the Tukey test for Honest Significant Differences), there is no effect of local complexity in complex landscapes, but in simple landscapes, locally complex sites have nearly four times the pest reduction as locally simple sites (p << 0.01). Likewise, there is no effect of landscape complexity at locally complex sites, but at locally simple sites, complex landscapes show a similar four-fold pest control advantage over simple landscapes (p << 0.01). It appears that local complexity can compensate for a lack of landscape complexity or vice versa. In terms of pest control function, sites with high local complexity and low landscape complexity are indistinguishable from sites with low local complexity and high landscape complexity. Caution must be exercised in the interpretation of this analysis due to the low sample size resulting from further breaking down the categories into local and landscape complexity. While 96 total cages stretched the limit of our logistical capacity and while four sites per landscape category was the maximum that could be acquired through farmer willingness to participate, adding the additional factor of local complexity reduces the effective sample size to 2 per group. Ameliorating this statistical concern may not be as straightforward as it sounds, however. Reducing within-site replication would compromise the integrity of the pest reduction measure, and changing the minds of farmers who were unwilling to participate will be difficult, considering their deep concerns about bringing pests (however well-contained) onto their farm. Therefore it may not be possible to increase landscape-level replication. Nonetheless, this experiment can and will be repeated next summer to determine whether the effects of local and landscape complexity are consistent across years. Laboratory experiments: physiological impacts of mustard
This component of the study is an important consideration in helping to explain the apparent lack of a relationship between aphids and landscape complexity over the whole gradient. Examining the middle of the landscape gradient for possible sources of aphids would be an important next step in understanding the role of landscape in aphid distributions; however, before aphid sources can be mapped, they must first be identified, and wild mustard is a plausible candidate for an aphid source due to its potential to serve as an enemy-free refuge for this pest.

The first question is whether aphids feeding on mustard contain more glucosinolates, the building blocks for the mustard bomb, than aphids feeding on broccoli. The answer to this is a definitive yes. The chemical assays showed that mustard-fed aphids contain more than ten times the glucosinolates of broccoli-fed aphids (Figure 10a, p < 0.01). While broccoli-fed aphids still have detectable levels of glucosinolates (as compared to the sugar-water-fed aphids in Kazana et al. 2007, which had none), it seems likely that the concentrations of these compounds are not high enough for aphids to build the mustard bombs to deter their enemies so effectively. The next question is whether syrphids are less effective predators of aphids on mustard as a result of this greater glucosinolate content. Initial results indicate that the answer to this is also yes. Syrphids ate almost five times more broccoli-fed aphids over their entire larval development than mustard-fed aphids (Fig. 10b, p < 0.01) and three times as much at their maximum (Fig. 10c, p < 0.01). But the most impressive result to come out of this study so far is the impact on larval development. Figure 11 shows average aphid consumption on a daily basis by syrphids on these two food sources, and the line for mustard clearly ends well before the line for broccoli. This is not because syrphids consuming mustard-fed aphids pupated more rapidly than syrphids on the broccoli-fed aphid diet, but rather, because no syrphids in the mustard treatment have survived to pupation at all! Syrphid mortality in the broccoli treatment was around 40%, while mortality in the mustard treatment is 100%. Although a syrphid larva is more or less confined to the plant it is born on (unless that plant is touching another, allowing the syrphid to move across), it seems likely that such high mortality rates would lead to fairly dramatic selection pressure against ovipositing on mustard. Rather than experiencing this kind of mortality in the field, it is possible that adult syrphid flies simply avoid mustard altogether. The final question is whether aphids trade off a physiological cost in exchange for building the mustard bomb. If aphids develop more slowly or reproduce less on mustard, it may counteract any potential effect of reduced mortality from predation, leading to a net neutral effect of mustard. If aphids are attracted to the mustard as a refuge from enemies but develop or reproduce more slowly there than when on broccoli, mustard could even be considered a sink rather than a source of aphids, and could be drawing pests away from the crop. If, on the other hand, aphid development and/or reproduction are not hindered by mustard and aphid population growth on mustard is less impacted by mortality from predation, then mustard could indeed be serving as a source of aphids to crops. The aphid development studies are underway, but have not produced results yet. However, initial observations indicate that another consideration may come into play besides length of development until reproduction and total lifetime fecundity: production of winged morphs. Aphids generally trigger production of winged morphs through hormonal signals in response to over-crowding, giving birth to nymphs that will develop into winged adults that can seek out plants with fewer aphids to begin a new colony. Poor food quality can also serve as signal to produce winged morphs, since over-crowding usually reduces food quality of the plant (through depletion of important nutrients or stressing the plant to the point of wilting). Initial observations suggest that aphids on mustard may be producing more winged aphids than aphids on broccoli, perhaps in response to the higher glucosinolate content of mustard that may render it lower quality food in the eyes of aphids. This could mean that even with slower development or reproduction, mustard could be producing a greater total number of winged individuals than broccoli and thus serve as a greater exporter of aphids. This possibility will need to be taken into account when mustard’s potential as a source of aphids is assessed. (4) and (5) Kremen C. and R. Chaplin-Kramer. 2007. Insects as providers of ecosystem services: crop pollination and pest control. In Insect Conservation Biology: proceedings of the Royal Entomological Society's 23rd Symposium. (Stewart, A.J.A., New, T.R. and Lewis, O.T. (eds)) CABI Publishing, Wallingford, 349-382.