SEARCH

SEARCH BY CITATION

Keywords:

  • agricultural intensification;
  • breeding productivity;
  • brood reduction;
  • chick diet;
  • chick growth rates;
  • indirect effects of insecticides;
  • population decline

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    The UK population of yellowhammers has declined since the mid-1980s. Concurrent increases in the use of pesticides are believed to have reduced the availability of food resources for farmland birds, including yellowhammers. To mitigate the consequences of insecticide applications on yellowhammer productivity, the relationships between insecticide application, arthropod food abundance and breeding success require quantification.
  • 2
    We studied nesting yellowhammers on a lowland arable farm in North Yorkshire between 2001 and 2003, to examine the effects of food abundance on breeding success and the effects of insecticide on food abundance. Arthropod abundances around individual nests were sampled and the timing and location of insecticide applications were recorded.
  • 3
    Nestling condition and mass on day 6 after hatching were positively correlated with the abundance of arthropods important in the diet of nestling yellowhammers. Greater mean body mass and condition corresponded with a lower incidence of brood reduction.
  • 4
    The abundance of arthropods important in the diet of nestling yellowhammers increased between mid-May and the end of July. However, arthropod samples collected within 20 days of an insecticide application did not show this seasonal increase in abundance and were depressed at levels likely to affect yellowhammer breeding performance adversely.
  • 5
    Synthesis and applications. We have demonstrated how insecticide applications can depress yellowhammer breeding productivity. We provide the requisite data for a framework that enables predictions to be made about the probable population effects of particular pesticide products. If the risk of indirect effects can be predicted accurately then appropriate mitigation and compensation measures could be incorporated into pesticide regulatory procedures and/or agri-environment schemes.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The populations of many farmland birds, including the yellowhammer Emberiza citrinella L., have undergone major declines in Britain (Newton 2004). Several hypotheses have been proposed to explain the declines. None the less, decreases in the abundance of weeds (Firbank & Smart 2002), seeds (Robinson & Sutherland 2002) and arthropods (Ewald & Aebischer 1999) on farmland in lowland Britain are thought to be significant drivers of recent change, acting through a combination of survival rates (Siriwardena, Baillie & Wilson 1998; Siriwardena & Robinson 2002) and reproductive output (Siriwardena et al. 2000). The extent to which pesticides have contributed to the population declines by reducing the food supplies of birds to levels that impact on survival or breeding productivity is unclear. There are three possible routes by which this may occur: insecticides reducing arthropod food supplies (type 1); herbicides eliminating plants that are hosts for arthropods taken by farmland birds (type 2); herbicides reducing the abundance of weeds, which provide either green matter or seeds for herbivorous and seed-eating species, respectively (type 3).

In principle, these processes do not involve direct poisoning of birds and have been termed indirect effects of pesticides (Newton 1995; Burn 2000). Indirect effects may be additive. For example, seed-eating passerines, which rely upon arthropod food supplies in summer to provision their young, could be vulnerable to all three types.

As yet, few examples of indirect effects of pesticides have been demonstrated experimentally in the field, mainly because of the laborious data requirements. In Britain, only the long-term study of the grey partridge Perdix perdix L. has amassed enough data to establish links between pesticide applications, arthropod food supplies, chick survival and population change. Type 1 and type 2 effects have been demonstrated (Rands 1985; Potts 1986). Type 3 effects are suspected to have contributed to the declines of seed-eating birds (Newton 1995) but none has yet been demonstrated empirically, partly because of analytical difficulties (Boatman et al. 2004). Nevertheless, indirect effects of pesticides have been identified as possible agents of the decline of 18 farmland bird species in Britain (Campbell et al. 1997; Boatman et al. 2004).

In this study we focused on type 1 effects by considering the impact of summer applications of insecticides on the arthropod food and reproductive output of the yellowhammer. The yellowhammer was chosen because it is a declining species that is potentially vulnerable to type 1 effects (Morris et al. 2005) and because it is still sufficiently abundant to provide adequate sample sizes for nest-based studies (Bradbury et al. 2000). The diet of yellowhammer nestlings is dominated by arthropods (although unripe grain may also be consumed late in the breeding season; Morris et al. 2005) and the adults forage in crops where their food supplies may be affected by insecticides (Stoate, Moreby & Szczur 1998; Morris, Bradbury & Wilson 2002).

In a meta-analysis of yellowhammer studies from eight farms in central and eastern England, Morris et al. (2005) presented evidence for some but not all indicators of a type 1 indirect effect. Arthropod food of yellowhammers was less abundant in cereal fields that had been sprayed with insecticides in summer than in those that had only been sprayed in winter or not at all. When foraging, adults tended to avoid fields that had received insecticide applications in summer, unless cereal grain was available. A quadratic relationship between chick condition and the number of insecticide applications in adjacent fields was also demonstrated. Chick condition was poorest in nests adjacent to fields that had received three applications, the third application usually being made in summer.

Morris et al. (2005) could not show a relationship between insecticide applications and chick starvation. However, in a replicated farm-scale experiment, in which summer insecticide inputs were deliberately increased in a proportion of fields to vary the extent of spraying around individual yellowhammer nests, Boatman et al. (2004) were able to demonstrate an effect. Where insecticides had been applied within 20 days of the hatch date, the probability of brood reduction was positively correlated with the proportion of the 200-m foraging range around each nest that had been sprayed. The most likely explanation for the observed brood reduction was chick starvation as a result of food shortages induced by the insecticides. Lethal or sublethal poisoning of chicks or provisioning adults was very unlikely because the insecticides used in the experiment were mostly pyrethroids (Cypermethrin, Deltamethrin and Lambda Cyhalothrin), which have low toxicity to birds (Anonymous 1989, 1990a,b), and because pyrethroids, unlike organophosphorus and carbamate products (Grue, Hart & Mineau 1991), are not known to impair the breeding and foraging behaviour of birds by inhibiting cholinesterase activity in the brain.

The relationship between the extent and timing of insecticide applications and the probability of brood reduction (Boatman et al. 2004) implied the existence of relationships between insecticide application, arthropod abundance, chick development and fledging success. This study aimed to quantify the relationships between food abundance in the foraging range of individual yellowhammer nests and chick development, and between chick development and the probability of brood reduction as a result of chick starvation. A further aim was to demonstrate that insecticide applications depressed arthropod abundance to levels where impairment of chick development, and chick loss, was likely.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

study area

The study area was located in Ryedale, North Yorkshire, UK. It comprised four c. 100-ha blocks of arable farmland. These blocks were used in a replicated field experiment in which arthropod food abundance was manipulated by varying insecticide inputs, to assess the scale of the indirect effects (Boatman et al. 2004). In 2002 arthropod food resources in two blocks were depressed by increasing insecticide inputs above normal practice. In this study all the blocks were used during 2001 and 2002, but only one in 2003. The percentage composition of crops by area are given for each year in Table S1 (see the supplementary material). Most field boundaries comprised either hedgerows or a combination of hedgerow and ditches covered by grasses and perennial forbs. Some fields also bordered woodland, dirt tracks and farm buildings. No unsprayed buffers or conservation headlands were used in any of the fields.

The hedgerows and ditches provided suitable nesting habitat for the yellowhammers. Male yellowhammers defended breeding territories along those field boundaries that contained suitable nesting habitat but they did not defend the surrounding crops that lay within the foraging range of their nests. The densities of breeding territories within the farmland blocks varied from 10·0 km−2 to 35·6 km−2 and were not atypical for farmland in Britain (Kyrkos, Wilson & Fuller 1998)

nest recording

Nests were monitored in 2001, 2002 and 2003, between late April and August. Nests were located by observing adult behaviour and by searching systematically in suitable locations. The progress of each nest was monitored by regular checks every 2–4 days but not daily, in order to reduce disturbance.

The outcome of each nest was classified according to whether the brood was predated, starved, deserted or fledged. Broods were assumed predated if the nest was found empty before the brood had reached a minimum fledging age of 8 days (J. D. Hart, unpublished data) or there was evidence that the nest had been disturbed. Nests found empty but undisturbed after the minimum fledging period were assumed to have fledged. Confirmation of fledging was obtained either by registering adult warning calls or observing provisioning behaviour in the vicinity of the nest. Partial losses of nestlings (brood reduction) were observed in some nests and it was assumed that these losses were the result of starvation. If the whole brood perished through repeated partial losses of young, it was assumed to have starved. The smallest and least developed chicks within a brood were assumed to be the first to perish from starvation (Hussell 1985). If a whole brood was found dead in the nest, with no prior evidence of starvation, it was assumed that the brood had been deserted.

measuring breeding performance

We assessed breeding performance by weighing nestlings to derive measures of chick development, and by recording the outcome of each nest.

Measurements of chick mass were made at 72 nests (29 broods in 2001, 34 broods in 2002 and nine broods in 2003). The masses of each chick were measured on two occasions between hatching and fledging (mean age of first weighing was 3·4 days, range 2–7 days, mean age of second weighing was 6·5 days, range 4–9 days). Whenever possible, measurements were confined to the first week after hatching. Growth rates were linear up to 7 days old and began to decline thereafter; therefore any effects of food depletion on growth should be amplified at this time (Bradbury et al. 2003). Chick tarsus length was also measured with the second mass measurement. Mass was measured to 0·1 g using a Pesola spring balance and tarsus length to 0·1 mm using dial callipers (method B of Svensson 1992). Mean growth rates per brood were calculated using the change in mean mass recorded between the first and second weighing dates. The difference was divided by the number of hours between the two measurements and then multiplied by 24 to give a daily growth rate (g day−1). A measure of mean chick condition was also derived using the ‘residual index’ from a regression that used mean tarsus length as a predictor of mean chick mass (Jakob, Samuel & George 1996); both variables were log-transformed to achieve homoscedasticity.

invertebrate sampling

Invertebrates were sampled in the foraging ranges of 25 nests containing broods (16 and nine broods in 2002 and 2003, respectively). Fourteen samples were collected from the foraging range around each nest. Sampling was stratified so that four samples were collected from the boundary strip between the crop margin and the hedgerow or ditch bordering the field (as defined by Greaves & Marshall 1987), and the remaining 10 from the surrounding crops. The crop samples were taken randomly from within a circle with a radius of 100 m, centred on the nest. Morris (2001) found that about 60% of yellowhammer foraging trips were within 100 m of the nest, and 80% within 200 m. Sampling was restricted to within a 100-m radius of the nest because logistically this was more simple and evidence suggested that food abundance within either a 100-m or 200-m radius from the nest could affect yellowhammer productivity. The linear relationship between the proportion of foraging area sprayed with insecticide and the probability of brood reduction demonstrated by Boatman et al. (2004) had similar intercepts and slope effects whether the foraging ranges were given a 100-m or 200-m radius around each nest. (Table S2, see the supplementary material).

The samples were collected using a Tecumseh Dietrick (D-Vac, Ventura, California, USA) insect suction sampler with a 35-cm diameter aperture on the sampling hose (Dietrick 1961). At each sample point, five subsamples were collected, with the suction hose of the D-Vac placed on the ground for a 5-second bout. These subsamples were combined to form one sample that was frozen for 12 h prior to storage and processing.

As invertebrate abundance is subject to large fluctuations over time, sampling was synchronized with the chick period. Samples were collected when the chicks of each nest were between 5 and 8 days old. All samples, for a given nest, were collected on the same day between 14:00 and 16:00 h, to minimize the effects of diurnal variation in invertebrate activity and abundance. Wet conditions were avoided because the efficiency of the D-Vac was reduced.

The samples collected from within the crops were separated from those collected from the field boundaries, and all the invertebrates counted and identified into the taxonomic groups given in Table S3 (see the supplementary material). A chick food index was calculated based on taxonomic groups known to be important in the diet of yellowhammer chicks (Stoate, Moreby & Szczur 1998; Moreby & Stoate 2001; Table S3, see the supplementary material). The taxonomic composition of the chick food index was checked by comparison with the contents of faecal samples collected from some of our broods. Faecal sacs were obtained when the chicks were handled to collect biometric data. A sample consisted of all the faecal sacs produced by a brood during a nest visit. The samples were preserved in 70% industrial ethanol and examined separately. Diet was determined following the methods outlined in Moreby & Stoate (2001).

data analysis

The relationships between chick food abundance and chick mass were investigated using linear regression analyses in Brodgar version 2·4·1 (http://www.brodgar.com). Using data obtained from the foraging ranges of 25 nests, we analysed the chick food counts from the cropped areas surrounding each nest.

Siblings within broods were not treated independently in the statistical analyses because they were fed by the same parents and experienced the same ambient conditions. We therefore calculated the mean mass of each brood and analysed the variation in mean chick mass of each nest as a function of the predictor variables. Some of the nests suffered brood reduction before any mass measurements were recorded or between the first and second set of chick mass measurements. Under these circumstances brood reduction could complicate the analyses because (i) when a chick dies the mean chick mass may be inflated relative to the unreduced brood (assuming the smallest chicks were more susceptible to starvation); and (ii) reduced sibling competition may also cause a change in the growth rates of the remaining chicks. After the loss of a sibling through starvation the remaining chicks in a brood can regain condition and normal growth rates (Shkedy & Safriel 1992). Nevertheless, we did not omit the reduced broods from the analyses as the incidence of brood reduction and its effect on chick development and survival may have been related to the food supply and be an integral part of the response to food abundance. Omitting the broods that lost siblings by starvation could have removed from the data set those nests that were most likely to have had chick masses reduced by food shortages.

Referring to Fig. S1 (see the supplementary material), it can be seen that broods in which some chicks died of starvation grew more slowly but the difference in growth rates between the reduced and unreduced broods was partly dependent on when the chicks were weighed in relation to the timing of brood reduction. The mean growth rates of broods that subsequently lost siblings as a result of starvation, after they had been weighed, were lower than the mean growth rates of the broods that lost no siblings (as would be expected from broods showing the effects of starvation). However, broods that had lost siblings by starvation before they were weighed had similar mean growth rates to the unreduced broods, suggesting that after the loss of a sibling the remaining chicks in a brood regained normal masses. However, most broods that suffered brood reduction lost siblings between their first and second mass measurements. These broods showed highly variable growth rates, which were on average significantly lower than the unreduced broods, but were likely to vary depending on the time and ability of the surviving chicks to compensate before the second weighing date. The occurrence of brood reduction potentially confounded any effects of food abundance on chick growth; we therefore treated brood reduction as a binary categorical factor in the analyses, allocating broods that had lost siblings before they were weighed with the unreduced broods.

We assessed the effects of chick food abundance and the confounding effect of brood reduction on chick mass using a multimodal inference approach based on Akaike's information criterion (AIC). Linear regressions were constructed using mean tarsus length as a predictor of chick mass, with the chick food counts and the occurrence of brood reduction included as covariates in separate models. Mean chick mass and tarsus length were log-transformed to achieve homoscedasticity. The tarsus lengths were included to allow for any effects of body size (mainly because of the variation in age when broods were measured; García-Berthou 2001). All possible subsets of the predictor variables (without omitting tarsus length) and their interaction terms were assessed using AIC corrected for small sample size (AICc). The candidate models were ranked according to their AICc scores and Akaike weights (w) used to identify the model that best explained our data (Burnham & Anderson 2002).

The relationships between chick mass and the probabilities of brood reduction and fledging success were modelled using binary logistic regression analysis in GenStat® (Payne et al. 2003). For these analyses we used data from the 38 broods that were weighed on day 6 after hatching between 2001 and 2003. This ensured that the masses were compared over the same phase of chick development.

insecticide inputs

In 2001 and 2003, pyrethroid insecticides were applied under normal farming practice. During 2002 insecticide inputs were increased above normal practice as part of a large-scale field experiment to depress food abundance and demonstrate the magnitude of any indirect effects (Boatman et al. 2004). Additional insecticide applications were applied to two of the farmland blocks.

Boatman et al. (2004) demonstrated that the probability of brood reduction was correlated with the proportion of the foraging range around nests that had been sprayed ≤ 20 days before hatching. We examined the effect of the time interval between an application of insecticide and the invertebrate D-Vac suction sampling date (see above) on the abundance of ‘chick food’ through the breeding season under two scenarios: (i) samples collected from fields that had been sprayed with insecticide within 20 days of the sample date; and (ii) samples from fields that had either been sprayed more than 20 days before the sample date or had had no prior application of insecticide. We modelled the effect of the spraying scenario using a GLM with a negative binomial distribution for error, and a logarithmic link function, fitted in GenStat® (Payne et al. 2003). The aggregation parameter for the negative binomial distribution was estimated by maximum likelihood before fitting the model (k = 1·4769). Terms for crop type and year were dropped from the model because adjusting for these variables did not result in a significant improvement in the fit of the model.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

chick diet composition

The mean proportions of each invertebrate group present within the faecal samples (Fig. 1a,b) suggested chick diet was composed mostly of the chick food group. Compositional analyses using the faecal samples and D-Vac-collected invertebrate samples also suggested that the chick food group of invertebrates were being preferentially selected (Wilks’ lambda F7,12= 51·378, P < 0·001; Aebischer, Robertson & Kenward 1993). Ranking the invertebrate groups suggested the following preferences: Lepidoptera > Coleoptera > Diptera, Arachnid, Dermaptera, Neuroptera > Hemiptera, Hymenoptera, other invertebrate groups.

image

Figure 1. The mean proportions of invertebrate orders present in the faecal samples of yellowhammer chicks collected during (a) 2002 (n = 48) and (b) 2003 (n = 12).

Download figure to PowerPoint

The Diptera, Arachnids, Dermaptera and Neuroptera did not differ in rank but these groups all ranked higher than the Hemiptera, Hymenoptera and other invertebrate groups that were also equally ranked.

effect of arthropod abundance on chick development

Using a multimodal inference approach based on AIC, we were unable to select a definitive model that best explained the variation in chick mass (adjusted for age using tarsus length) from our chosen predictor variables. The evidence that the top-ranked model was the best approximating model could be judged by the ratio of the Akaike weights (w) of the first and second ranked models (Burnham & Anderson 2002). This evidence (w1/w2 = 1·03) was insufficient for us to ignore model selection uncertainty. The top three ranked models were the best supported (with Δi = 2). These models had a summed Akaike weighting (w) of 0·78 and were composed of tarsus length and chick food abundance with and without the occurrence of brood reduction and their interaction term as predictors of chick mass (Table 1). Empirical support for an effect of chick food was greater than for an effect of brood reduction; the summed Akaike weightings for the models that included these covariates were 0·78 and 0·57, respectively. The effect of chick food was significant in all the models, despite evidence that the occurrence of brood reduction after the first weighing date also affected mean chick mass at the second weighing date (Table 1). Overall, the model-averaged estimate of the chick food effects on chick mass was significantly different from zero (using 95% confidence intervals, ± 2 SE; Table 1). For illustrative purposes, Fig. 2a shows the effect of chick food abundance on chick mass (adjusted for age using tarsus length) for the model that omitted brood reduction; the slope and intercept for this model were similar to the model-averaged estimate (Table 1).

Table 1.  Results of AIC-based model selection. The table shows the AICc scores, the number of model parameters (K), the differences between the lowest AICc scored model and each candidate model (Δi), the Akaike weights (wi) and the estimate of the chick food effect on chick mass (CF effect) for each candidate model and their estimated conditional standard error, SE (effect)
RankData set 1KAICcΔiwiCF effectSE (effect)
  1. TS, tarsus length (log10); CFI, chick food count; BR, brood reduction occurrence; NA, not applicable.

1TS, CFI, BR3−162·61200·330·000300·00014
2TS, CFI2−162·5160·0960·320·000320·00015
3TS, CFI, BR, CFI × BR4−160·6851·9270·130·000410·00019
4TS, BR2−160·4152·1970·11NANA
5TS1−160·4132·1990·11NANA
Model averaged    0·000250·00012
image

Figure 2. The relationships between chick food abundance, chick mass and the occurrence of brood reduction. (a) The fitted line of the chick food effect from a regression model that used tarsus length and chick food abundance as predictors of chick mass (slope = 0·00032, SE = 0·00015, t22 = 2·16, P < 0·05; R2 = 90·0%, F2-22 = 109·51, P < 0·001); and (b) the fitted line from a binary logistic regression showing the relationship between mean chick mass on day 6 after hatching and the probability of brood reduction (slope = −0·560, SE = 0·201, t37 = − 2·78, P < 0·01; deviance ratio = 14·76, d.f. = 1, P < 0·001).

Download figure to PowerPoint

chick development and nest outcome

Of 72 nests that hatched during 2001–03, 54 broods fledged at least one chick, five broods starved and 10 broods suffered predation (three broods had unknown outcomes). Broods with lower mean weights (measured on day 6 after hatching) were more likely to have suffered brood reduction or to have lost siblings to starvation since the first weighing date (Fig. 2b). The probability of brood reduction was greater than 0·5 where chicks weighed on average less than 14·0 g (log10 = 1·15) on day 6 after hatching. Mean chick condition could also predict the likelihood of brood reduction (Fig. 3a).

image

Figure 3. The fitted lines from binary logistic regressions showing the relationships between (a) the probability of brood reduction and a measure of chick condition based on the residuals of a regression of mean chick mass and tarsus length (slope =−0·710, SE = 0·336, t65 = −2·11, P < 0·05; deviance ratio = 5·17, d.f. = 1, P < 0·05); and (b) the probability of fledging and mean chick mass on day 6 after hatching (slope = 9·19, SE = 4·25, t37 = 2·17, P < 0·05; deviance ratio = 6·90, d.f. = 1, P < 0·01).

Download figure to PowerPoint

Mean chick mass (measured on day 6 after hatching) was also a good predictor of fledging (Fig. 3b). The probability of fledging increased from 0·4 for broods that weighed on average 10·0 g (log10 = 1·0) to more than 0·8 for broods that weighed > 16·0 g (log10 = 1·20). However, the significant relationship between chick mass and fledging success was influenced by the number of nests that lost their whole brood to starvation. If we included only fledged and predated broods, excluding starved broods to determine whether predation risk could be predicted from chick mass, then mean chick mass could no longer be used to predict fledging success (binary logistic regression, deviance test = 4·27, d.f. = 1, P < 0·05, slope = 8·10, SE = 4·54, t36 = 1·78 NS, deviance test = 4·27, d.f. = 1, P < 0·05). Therefore, the probability of nest failure as a result of predation could not be predicted from mean chick mass (measured on day 6 after hatching).

arthropod abundance and the effect of insecticide

The abundance of invertebrates was greater in the boundary strips than within the crops throughout May–July. During May and June, chick food was also more abundant in the boundary strips but, by July, its abundance had increased markedly in the crops and there was no longer a significant difference between the strips and the crops (Table 2a). By July, there was a significant increase in the proportion of chick food found in the invertebrate samples collected from the crops (anova, F2,289 = 20·71, P < 0·001). This change in proportion was not observed from samples collected in the boundary strips (F2,63 = 1·91, NS; Table 2b).

Table 2a.  A comparison between the boundary strip and surrounding crops of the mean total invertebrate (TOT) and chick food (CFI) counts (mean ± SE) from within each nests’ foraging range during May, June and July. Paired t-tests were possible only for June (n = 7) and July (n = 6) because the sample size was too small in May (n = 3)
 CropField edgePaired t-test
TOTCFITOTCFITOTCFI
  1. P < 0·05, *P < 0·01, **P < 0·001.

May147·9 ± 29·2 45·7 ± 1·5558·8 ± 270·8147·1 ± 28·5NANA
June175·8 ± 28·2 59·8 ± 6·9620·6 ± 164·5169·3 ± 36·8−3·08*−3·17*
July265·5 ± 32·1117·3 ± 13·6420·9 ± 18·6166·0 ± 10·6−4·18**−1·23 NS
Table 2b.  The mean proportion of each invertebrate sample collected within the crops and the boundary strips that was composed of chick food during May, June and July
MonthCropField edge
Proportion (mean ± SE)Sample size (n)Proportion (mean ± SE)Sample size (n)
May0·3525 ± 0·0289 300·3515 ± 0·043112
June0·3623 ± 0·01431200·3392 ± 0·029948
July0·4826 ± 0·01411400·3215 ± 0·025856

The counts of chick food per D-Vac sample point were reduced in fields that had received an application of insecticide. This effect, however, was only apparent from samples collected from the treated crops within 20 days of an insecticide application. The model in Table 3a shows that the insecticide effect increased between mid-May and mid-July. Between these dates, numbers of chick food items in unsprayed samples from the crops showed a significant tendency to increase, whereas the sprayed samples remained depressed (Table 3a and 3b andFig. 4).

Table 3a.  The sequential deviance table for the GLM for the effects of spraying regime and sampling date on the chick food count per sample
 d.f.DevianceMean devianceDeviance ratioApproximate χ-square probability
Spray regime  1 10·3810·3810·38   0·001
Sample day  1 60·9660·9660·96< 0·001
Interaction  1  6·86 6·86 6·86   0·009
Residual286239·97 0·839  
Total289318·17 1·101  
Table 3b.  Parameter estimates for the two linear equations
 EstimateSE
Unsprayed
Constant1·6140·341
Slope0·024320·00289
Sprayed
Constant3·0790·551
Slope0·009610·00472
image

Figure 4. The effects of sample date and the application of insecticide on the count of ‘chick food’ per sample within the cropped portion of fields. Day 80, 19th May; sprayed (black circles with solid trend line), samples collected from fields < 20 days after an application of insecticide; unsprayed (open circles with dashed trend line), samples that were collected from unsprayed fields or fields that had been treated with insecticide > 20 days before the sampling date.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Relationships between the abundance of arthropod prey that are available to provisioning adult yellowhammers, chick development and the likelihood of brood reduction have been quantified. They have been reduced to two interlinked regression models with which it is possible to predict the probability of yellowhammer chick losses occurring, using a standardized measure of food abundance around the nest. Over the short-term, insecticide applications made in summer consistently depressed the abundance of arthropods, which are important in the yellowhammer diet, to levels where reductions in chick condition and fledging success were likely. We suggest that this process underlies the relationship between the timing and extent of insecticide application and brood reduction reported by Boatman et al. (2004).

mechanisms

The discovery of a relationship between chick condition and arthropod food abundance implies that, on average, pairs did not fully compensate for food shortages around the nest. If arthropod numbers were low then this limited chick growth. The variability in chick mass (adjusted for age using tarsus length), at given food abundances, may reflect the plasticity in the provisioning behaviour of the adults. At nests where food abundance was low and the mean chick condition was higher than average, the provisioning adults may have attempted to compensate for food shortages by working harder when foraging in areas where prey abundance had been depressed, by exploiting alternative food resources, such as unripe cereal grain (Morris et al. 2005), or by foraging in habitats other than the crops. However, in the case of pairs with nests that were completely surrounded by recently sprayed fields, the energetic constraints of central place foraging (Stephens & Krebs 1986) may have rendered alternative foraging patches outside the spray area unprofitable. Variation in parental investment may also have contributed to the variability in chick growth. For example, the provisioning effort of each sex varied between nests; some male yellowhammers did not provision at all while their nests were under observation (J. D. Hart, unpublished data).

Yellowhammer chicks are not fed invertebrates exclusively. By mid-summer, some broods are also fed unripe cereal grain taken from fields that have been sprayed recently with insecticide and where arthropod numbers may be low (Morris et al. 2005). Although cereal grain is easily collected, the energy contents (kJ g−1) of barley Hordeum sp. and wheat Triticum sp. are both lower than that of arthropods (Christensen, Falk & Petersen 1996), as is the protein content of wheat (Potts 1986). The low protein content of grain is likely to be a limiting factor because chicks require food that is rich in protein and specific amino acids, which may be lacking in plant food (Ricklefs 1983; Potts 1986). We could not determine the value of grain as a buffer to shortages of arthropods induced by insecticides, but chicks require protein to grow and our results imply arthropods are a necessary source.

Season differences in the abundance of chick food between the crops and field boundary strips indicate that the availability of alternative foraging habitats could affect chick development. During May and June, the counts of chick food in the boundary strips were, on average, greater than those in the adjacent crops (Table 2a). By July, however, chick food was no more abundant in the field edge than in the crop, and a greater proportion of the crop invertebrate population was also composed of chick food (Table 2b). Consequently, by late June and July the crops potentially provided as good a source of food as the field boundary habitats. Yellowhammer broods reared where field boundaries provided poor foraging habitat, and broods reared later in the season, when dense plant growth reduces accessibility to the field boundaries, may be more reliant on chick food collected from the surrounding crops. These broods would be more vulnerable to changes in the food abundance within the crops and therefore also to any indirect effects of insecticide.

The analyses have focused upon chick death from starvation. However, an increased risk of brood loss as a result of predation may be a secondary effect of food shortages induced by insecticides. It has been proposed that hungry broods are at greater risk of predation than broods that are well nourished because hungry chicks attract the attention of predators by begging more loudly and for longer (Haskell 1994) and because low invertebrate abundance increases the time parents spend away from the nest (Brickle et al. 2000). However, we found no evidence that broods with low chick mass (on day 6 after hatching) were more likely to be predated than those with high masses.

Many altricial species are unable to maintain chick mass during food shortages (Konarzewski et al. 1996). Our results suggested that after the loss of a sibling through starvation, the remaining chicks in a brood had similar mean growth rates to the unreduced broods (Fig. S1, see the supplementary material) but mean chick mass (recorded at the final weighing date and adjusted for age) was lower when food was less abundant (Fig. 2a). Yellowhammer chicks, like most open-nesting small passerines, are probably physiologically unable to slow their development during periods of food shortage. It is therefore likely that poorly fed yellowhammer broods will fledge under-weight, which may reduce survival and even reproductive fitness (cf. Newton 1998).

insecticide effects

It has been demonstrated that summer applications of insecticides will depress the abundance of yellowhammer chick food to levels that are predicted to have an adverse effect on chick development and fledging success (Fig. 4). The severity of the impact will depend upon the extent of applications around individual nests and their timing relative to the hatch date (Boatman et al. 2004). Where isolated fields are sprayed during the nesting period, yellowhammers provisioning their young may simply select unsprayed fields and field boundary habitats nearby (Morris et al. 2005). However, as spraying becomes more extensive and the choice of unsprayed fields becomes more restricted, there will be a concomitant decrease in the mean abundance of chick food in the foraging range to levels where chick starvation and brood reduction are likely. Although field boundaries provided a good source of chick food (Table 2a and 2b), the effect of extensive spraying may have intensified later in the season when dense plant growth reduced accessibility to the field boundaries, and adults may have become more reliant on chick food collected from the surrounding crops.

In the farm-scale experiment, some nests were completely surrounded by fields that had been recently sprayed and there is evidence that routine spraying in the past attained similar coverage. In the Defra (Department for Environment, Food and Rural Affairs) eastern region of England, where June applications of insecticides have been particularly extensive, the proportion of farms where more than 75% of fields by area were treated increased from 13% in 1988 to a peak of 25% in 1990 and then declined to 6·5% in 2002 (Pesticide Usage Survey, unpublished data). It seems likely that the impact of insecticide applications on the outcome breeding attempts by yellowhammers in June may have been extensive in the early 1990s.

The timing of insecticide applications in Great Britain varies considerably between crops but, overall, the frequency of applications peaks in autumn and, again, in early summer (Garthwaite et al. 2003). In this study, we have focused on summer applications because the variation in the incidence of brood reduction in yellowhammer nests was not related to applications made the preceding autumn (Boatman et al. 2004). This finding concurs with that of Morris et al. (2005), who showed that the impact of insecticides on the abundance of arthropods, which are important in the yellowhammer diet, was related to the timing of the treatments. Within growing seasons, the effects of winter insecticide applications were relatively unimportant compared with those of summer sprays.

As indirect effects are defined by their impact on populations, the reduction in either breeding productivity or survival must be of sufficient magnitude to reduce the population growth rate of the bird species affected. Currently, techniques for modelling the consequences of changing food supply are constrained by a lack of data (Stephens et al. 2003). Habitat-based models have linked farm management practices to population sizes (Whittingham et al. 2005) but reliable quantitative predictions about population level responses need to be based on processes operating at the individual level (Norris 2004). In a series of interlinked equations, we have quantified the relationships between insecticide applications, arthropod food abundance and yellowhammer productivity. These relationships could be used in a framework that enables predictions to be made about the likely population effects of particular pesticide products. If the risk of indirect effects can be predicted accurately then appropriate mitigation and compensation measures could be introduced either through pesticide regulatory procedures or agri-environment schemes. A framework for predicting indirect effects could assist ecological evaluations of agri-environment prescriptions designed to reverse the decline in farmland birds (cf. Kleijn & Sutherland 2003).

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The Pesticide Safety Directorate of Defra funded the work under Project PN0925.

We are very grateful to the landowners for generously granting permission to work on their estate, and to the farm staff for their co-operation. We are also grateful to the British Atmospheric Data Centre for providing us with meteorological data and thank Deborah Beaumont, Jo Marshall, Carl Wardill, Tim Drew, George Watola, Dave Parrott and Richard Walls for assisting in the collection and collation of field data. Additional thanks are due to Carola Deppe and Alain Zuur for statistical help, Nigel Boatman, Joe Crocker, Mark Clook and John Holland for useful discussion, and to Dan Chamberlain for comments on an earlier draft.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Aebischer, N.J., Robertson, P.A. & Kenward, R.E. (1993) Compositional analysis of habitat use from animal radio-tracking data. Ecology, 74, 13131325.
  • Anonymous (1989) Environmental Health Criteria 82: Cypermethrin. International Programme on Chemical Safety. World Health Organization, Geneva, Switzerland.
  • Anonymous (1990a) Environmental Health Criteria 97: Deltamethrin. International Programme on Chemical Safety. World Health Organization, Geneva, Switzerland.
  • Anonymous (1990b) Environmental Health Criteria 99: Cyhalothrin. International Programme on Chemical Safety. World Health Organization, Geneva, Switzerland.
  • Boatman, N.D., Brickle, N.W., Hart, J.D., Milsom, T.P., Morris, A.J., Murray, A.W.A., Murray, K.A. & Robertson, P.A. (2004) Evidence for the indirect effect of pesticides on farmland birds. Ibis, 146 (Supplement 2), 131143.
  • Bradbury, R.B., Kyrkos, A., Morris, A.J., Clark, S.C., Perkins, A.J. & Wilson, J.D. (2000) Habitat associations and breeding success of yellowhammers (Emberiza citrinella) on lowland farmland. Journal of Applied Ecology, 37, 789805.
  • Bradbury, R.B., Wilson, J.D., Moorcroft, D., Morris, A.J. & Perkins, A.J. (2003) Habitat and weather are weak correlates of nestling condition and growth rates of four UK farmland passerines. Ibis, 145, 295306.
  • Brickle, N.W., Harper, D.G.C., Aebischer, N.J. & Cockayne, S.H. (2000) Effects of agricultural intensification on the breeding success of corn buntings (Miliaria calandra). Journal of Applied Ecology, 37, 742755.
  • Burn, A.J. (2000) Pesticides and their effects on lowland birds. Ecology and Conservation of Lowland Farmland Birds (eds N.J.Aebischer, A.D.Evans, P.V.Grice & J.A. Vickery), pp. 89104. British Ornithologists’ Union, Tring, UK.
  • Burnham, K.P. & Anderson, D.R. (2002) Model Selection and Multi-Model Inference: A Practical Information-Theoretic Approach, 2nd edn. Springer-Verlag, New York, NY.
  • Campbell, L.H., Avery, M.I., Donald, P., Evans, A.D., Green, R.E. & Wilson, J.D. (1997) A Review of the Indirect Effects of Pesticides on Birds. JNCC Report No. 227. Joint Nature Conservation Committee, Peterborough, UK.
  • Christensen, K.D., Falk, K. & Petersen, B.S. (1996) Feeding Biology of Danish Farmland Birds. Arbejdsrapport fra Miljostyrelsen 12. Danish Environmental Protection Agency, Copenhagen, Denmark.
  • Dietrick, E.J. (1961) An improved backpack motorised fan for suction sampling of insects. Journal of Economic Entomology, 54, 394395.
  • Ewald, J.A. & Aebischer, N.J. (1999) Avian Food, Birds and Pesticides. JNCC Report no. 296. Joint Nature Conservation Committee, Peterborough, UK.
  • Firbank, L.G. & Smart, S.M. (2002) The changing status of arable plants that are important food items for farmland birds. Birds and agriculture. Aspects of Applied Biology, 67, 165170.
  • García-Berthou, E. (2001) On the misuse of residuals in ecology: testing regression residuals vs. the analysis of covariance. Journal of Animal Ecology, 70, 708711.
  • Garthwaite, D.G., Thomas, M.R., Dawson, A. & Stoddart, H. (2003) Pesticide Usage Survey Report 187: Arable Crops in Great Britain 2002. Department for Environment, Food and Rural Affairs, London, UK.
  • Greaves, M.P. & Marshall, E.J.P. (1987) Field margins: definitions and statistics. Field Margins (eds J.M.Way & P.W.Greig-Smith), pp. 8594. BCPC Monograph No. 35. BCPC Publications, Thornton Heath, UK.
  • Grue, C.E., Hart, A.D.M. & Mineau, P. (1991) Biological consequences of depressed brain cholinesterase activity in wildlife. Cholinesterase-Inhibiting Insecticides: the Impact on Wildlife and the Environment (ed. P.Mineau), pp. 151209. Elsevier, Amsterdam, the Netherlands.
  • Haskell, D. (1994) Experimental evidence that nestling begging behaviour incurs a cost due to nest predation. Proceedings of the Royal Society of London, Series B-Biological Sciences, 257, 161164.
  • Hussell, D.J.T. (1985) On the adaptive basis for hatching asynchrony: brood reduction, nest failure and asynchronous hatching in snow buntings (Plectrophenax nivalis). Ornis Scandinavica, 16, 205212.
  • Jakob, E.M., Samuel, D.M. & George, W.U. (1996) Estimating fitness: a comparison of body condition indices. Oikos, 77, 6167.
  • Kleijn, D. & Sutherland, W.J. (2003) How effective are European agri-environment schemes in conserving and promoting biodiversity? Journal of Applied Ecology, 40, 947969.
  • Konarzewski, M., Kowalczyk, J., Swierubska, T. & Lewonczuk, B. (1996) Effect of short-term feed restriction, realimentation and overfeeding on growth of song thrush (Turdus philomelos) nestlings. Functional Ecology, 10, 97105.
  • Kyrkos, A., Wilson, J.D. & Fuller, R.J. (1998) Farmland habitat change and abundance of yellowhammers (Emberiza citrinella): an analysis of common birds census data. Bird Study, 45, 232246.
  • Moreby, S.J. & Stoate, C. (2001) Relative abundance of invertebrate taxa in the nestling diet of three farmland passerine species, dunnock (Prunella modularis), whitethroat (Sylvia communis) and yellowhammer (Emberiza citrinella) in Leicestershire, England. Agriculture, Ecosystems and Environment, 86, 125134.
  • Morris, A.J. (2001) Assessing the Indirect Effect of Pesticides on Birds. RSPB Progress Report. Department for Environment Food and Rural Affairs, London, UK.
  • Morris, A.J., Bradbury, R.B. & Wilson, J.D. (2002) Determinants of patch selection by yellowhammers Emberiza citrinella foraging in cereal crops. Birds and agriculture. Aspects of Applied Biology, 67, 4350.
  • Morris, A.J., Wilson, J.D., Whittingham, M.J. & Bradbury, R.B. (2005) Indirect effects of pesticides on breeding yellowhammer (Emberiza citrinella). Agriculture, Ecosystems and Environment, 106, 116.
  • Newton, I. (1995) The contribution of some recent research on birds to ecological understanding. Journal of Animal Ecology, 64, 675696.
  • Newton, I. (1998) Population Limitation in Birds. Academic Press Limited, London, UK.
  • Newton, I. (2004) The recent declines of farmland bird populations in Britain: an appraisal of causal factors and conservation actions. Ibis, 146, 579600.
  • Norris, K. (2004) Managing threatened species: the ecological toolbox, evolutionary theory and declining-population paradigm. Journal of Applied Ecology, 41, 413426.
  • Payne, R.W., Baird, D.B., Cherry, M., Gilmour, A.R., Harding, S.A., Kane, A.F., Lane, P.W., Murray, D.A., Soutar, D.M., Thompson, R., Todd, A.D., Tunnicliffe Wilson, G., Webster, R. & Welham, S.J. (2003) GenStat® Release 7·1 Reference Manual. VSN International, Oxford, UK.
  • Potts, G.R. (1986) The Partridge: Pesticides, Predation and Conservation. Collins, London, UK.
  • Rands, M.R.W. (1985) Pesticide use on cereals and the survival of grey partridge chicks: a field experiment. Journal of Applied Ecology, 22, 4954.
  • Ricklefs, R.E. (1983) Avian postnatal development. Avian Biology (ed D.S.Farner, J.R.King & K.C. Parkes), Vol. VII, pp. 181. Academic Press, New York and London.
  • Robinson, R.A. & Sutherland, W.J. (2002) Post-war changes in arable farming and biodiversity in Great Britain. Journal of Applied Ecology, 39, 157176.
  • Shkedy, Y. & Safriel, U.N. (1992) Nest predation and nestling growth rate of two lark species in the Negev desert, Israel. Ibis, 134, 268272.
  • Siriwardena, G.M. & Robinson, R.A. (2002) Farmland birds: demography and abundance. Birds and agriculture. Aspects of Applied Biology, 67, 179188.
  • Siriwardena, G.M., Baillie, S.R., Crick, H.Q.P. & Wilson, J.D. (2000) The importance of variation in the breeding performance of seed-eating birds in determining their population trends on farmland. Journal of Applied Ecology, 37, 128148.
  • Siriwardena, G.M., Baillie, S.R. & Wilson, J.D. (1998) Variation in the survival of some British passerines with respect to their population trends on farmland. Bird Study, 45, 276292.
  • Stephens, D.W. & Krebs, J.R. (1986) Foraging Theory. Monographs in Behaviour and Ecology. Princeton University Press, Princeton, NJ.
  • Stephens, P.A., Freckleton, R.P., Watkinson, A.R. & Sutherland, W.J. (2003) Predicting the response of farmland bird populations to changing food supplies. Journal of Applied Ecology, 40, 970983.
  • Stoate, C., Moreby, S.J. & Szczur, J. (1998) Breeding ecology of farmland yellowhammers (Emberiza citrinella). Bird Study, 45, 109121.
  • Svensson, L. (1992) Identification Guide to European Passerines. Lars Svensson, Stockholm, Sweden.
  • Whittingham, M.J., Swetnam, R.D., Wilson, J.D., Chamberlain, D.E. & Freckleton, R.P. (2005) Habitat selection by yellowhammers (Emberiza citrinella) on lowland farmland at two spatial scales: implications for conservation management. Journal of Applied Ecology, 42, 270280.