Seasonal changes in the response of oystercatchers Haematopus ostralegus to human disturbance

5361 JOURNAL OF AVIAN BIOLOGY 33: 358-365, 22 Seasonal changes in the response of oystercatchers Haematopus ostralegus to human disturbance Richard A. Stillman and John D. Goss-Custard Stillman, R. A. and Goss-Custard, J. D. 22. Seasonal changes in the response of oystercatchers Haematopus ostralegus to human disturbance. - J. Avian. Biol. 33: 358-365. The response of foraging animals to human disturbance can be considered as a trade-off between the increased perceived predation risk of tolerating disturbance and the increased starvation risk of not feeding and avoiding disturbance. We show how the response of overwintering oystercatchers Haematopus ostralegus to disturbance is related to their starvation risk of avoiding disturbance. As winter progresses, oystercatcher energy requirements increase and their feeding conditions deteriorate. To survive they spend longer feeding and so have less spare time in which to compensate for disturbance. Later in winter, birds approach a disturbance source more closely and return more quickly after a disturbance. Their behavioural response to disturbance is less when they are having more difficulty surviving and hence their starvation risk of avoiding disturbance is greater. These results have implications for studies which assume that a larger behavioural response means that a species is more vulnerable to disturbance. The opposite may be true. To more fully understand the impact of disturbance, studies should measure both behavioural responses and the ease with which animals are meeting their requirements. Conservation effort should be directed towards species which need to spend a high proportion of their time feeding, but still have a large response to disturbance. R. A. Stillman (correspondence) and J. D. Goss-Custard, Centre for Ecology and Hydrology (CEH) Dorset, Winfrith Technology Centre, Winfrith Newburgh, Dorchester, Dorset DT2 8ZD, UK. E-mail: rast@ceh.ac.uk There is considerable debate into the effects of human disturbance on animal populations (e.g. Hockin et al. 1992, Davidson and Rothwell 1993, Hill et al. 1997). The response to disturbance is often measured as the distance over which animals respond to disturbance or the time to return after human activity has ceased (e.g. Smit and Visser 1993). An assumption in the interpretation of many of these studies is that disturbance is more serious (e.g. causes more animals to die) when the behavioural response to disturbance is greater (e.g. animals take longer to return after disturbance has stopped or are excluded from a larger area by disturbance) (Gill et al. 21). The response to disturbance can be considered as a trade-off between food intake and the perceived predation risk of human presence (Gill et al. 1996, Sutherland 1996). Animals choose either to tolerate human presence by continuing to feed, but at an increased perceived predation risk, or to avoid it and decreasing their perceived risk, but at the expense of reduced intake and hence increased starvation risk. A prediction of this framework is that the response to disturbance will depend on the starvation risk of avoiding the disturbance; animals may respond less when their starvation risk is greater (Gill et al. 21). We measured the response to disturbance of an overwintering shorebird, the oystercatcher Haematopus ostralegus. Shorebirds may be particularly vulnerable to disturbance as their feeding areas are frequently used by people, and they have high energy demands (Kersten and Piersma 1987), that become increasingly difficult to meet as winter progresses and feeding conditions deteriorate (Goss-Custard et al. 1996) and thermoregulatory energy demands increase (Wiersma and Piersma 1994). To survive they must spend longer feeding and so have less spare time in which to compensate for disturbance. C JOURNAL OF AVIAN BIOLOGY 358 JOURNAL OF AVIAN BIOLOGY 33:4 (22)

53611 We show how the response to disturbance varies as energy demands and feeding conditions change. Methods Experiments were performed between October 1994 and September 1996 on mussel Mytilus edulis bed 2 of the Exe estuary, England (see Goss-Custard et al. (1982) for a full description of the Exe estuary's mussel beds). This intertidal bed is near the shoreline, has an area of 9.4 ha, and is fully exposed on spring tides for about two hours. It is relatively undisturbed in comparison with other parts of the estuary (Goss-Custard and Verboven 1993), allowing the controlled distur- bances to be the only major disturbance during experiments. However, the bed is not completely undisturbed and birds encounter human activities (walking, bait collecting and shellfishing) on a regular basis; during 25% of experiments some birds on the bed were disturbed as a person either passed nearby or crossed part of the bed. Changes in the response to disturbance were measured by repeating experiments throughout winter. In order to reduce the likelihood of birds habituating to the disturbance, experiments were conducted infrequently, on successive spring tide series and so about 14 days apart. The experiments were also designed to replicate the types of human activity normally occurring on or near the bed. Experiments were performed on spring tides because only part of the bed was exposed on neap tides. Results were very similar each year, and so were combined for analysis. We measured the response to disturbance at two scales. The bed-wide disturbance measured the combined response of the population of birds on the bed, while the local disturbance measured the response of birds around a disturbance source. source of disturbance throughout winter. As soon as the person left the bed, the observer restarted counting at 1-2 min intervals until the bed was completely covered by the advancing tide, approximately 4 h later. Control counts were made of the number of birds on the bed throughout the exposure period on two undisturbed days in early and late winter. The response to disturbance was measured from the number of birds excluded from the bed and the time taken for them to return afterwards. The number of birds expected to be on the bed over low tide was calculated by modelling the changes in bird numbers from 9 min after first exposure to the time of disturbance (Fig. la). C -o E.3 (a) Bed-wide experiment...........n..x. 2 amx n r 1 15,t 1 2 3 Time since first exposure (t)(min) (b) Local experiment Bed-wide disturbance An observer, in an elevated position on the shoreline, counted the number of feeding and non-feeding birds on the bed and on an adjacent sand ridge (on which birds were known to gather after disturbances) at 1-2 min intervals from first exposure of the bed on the receding tide. Counts continued for about two hours until the bed was fully exposed. At low tide a second person walked across the bed, following a standard route so that all parts of the bed were approached to within approximately 75 m and all birds took flight. In most experiments the majority of birds left the bed and settled on or near the sand ridge, but in late winter many birds returned immediately to the bed, landing on areas previously disturbed. In these cases the person did not alter his route to re-disturb areas of the bed, but adhered to the standard route to ensure a constant t- C1 O 1 5 25 5 75 1 125 Time since start of disturbance (min) Fig. 1. The response of oystercatchers to disturbance. (a) Bed-wide experiment - observed and fitted bird numbers during a day with disturbance; (b) local experiment - exclusion distance (mean + standard error) against time since the disturbance started. In (a), numbers prior to the disturbance are described by equation 1, numbers after by equation 2 and the amount of lost feeding time (min bird - ) measured from the shaded area A. JOURNAL OF AVIAN BIOLOGY 33:4 (22) 359

53612 nmaxi m1 - r(ti - t) if t < t n = nmaxi if t > t1 where n = number on bed, t = time since first exposure (min), nmaxl = maximum number on bed at low water, t, =time at which numbers reach a maximum and r = rate at which numbers increase. Non-linear regression was used to estimate parameters for each of the experimental days. This model accurately described the build up of birds on the bed for all replicates (mean r2 = 97.8%, range = 91.5-99.4%). As the control counts showed that the number of birds on the bed remained relatively constant while it was fully exposed (see below), the predicted value of n immediately prior to disturbance was assumed to be the number that would have occupied the bed during low tide if the disturbance had not occurred. After the disturbance, but before the tide started to cover the bed, the return of birds was modelled using a similar approach (Fig. la). n max2 (1) - g(t2 - t) if t < t2 (2) nmax2 if t > t2 where nmax2 = maximum number of birds on bed after disturbance, t2 = time at which number of birds on bed reaches a maximum and g = rate at which birds reoccupy bed. Non-linear regression was used to estimate parameters for each of the experiments, and the model accurately described the changes in bird numbers (mean r2 = 91.4%, range = 44.7-98.9%). The bed-wide return time was measured as the amount of lost feeding time per bird during the 45 min after the disturbance (Fig. la). Local disturbance The local disturbancexperiments were performed on a 25 x 15 m, relatively level (3 min from first to complete exposure) part of the mussel bed, located about 25 m from, and running perpendicular to the shoreline. To measure the location of birds, this transect was divided into six 25 x 25 m cells using white plastic markers (approximately 15 x 15 cm in cross section and 25 cm high) which remained in location throughout the study and were visible from the shore. Experiments started when the bed was fully exposed and finished when the tide started to cover the bed, approximately two hours later. After the bed was fully exposed and before the disturbance, an observer, located on the shore with a view along the longest length of transect, counted the number of birds in each cell at 5-min intervals and the numbers of feeding and non-feeding birds throughout the bed at approximately 2-min intervals. The observer then walked on to the bed, directly to the transect, until he reached the nearest edge of the nearest cell (as the transect ran perpendicular to the shore and the observer's initial position on the shore was always the same, the route to the transect and the final location of the observer on the transect were the same in each replicate experiment). Once at the transect, the observer counted the number of birds in each cell every 5 min. During the first winter of the study, the observer remained in place (for up to 2 h) until the advancing tide started to cover the bed. In subsequent winters, the observer remained on the bed for approximately 3 min and then returned to the initial location on the shoreline. Once on the shore, the number of birds in each cell was again counted every 5 min until the advancing tide started to cover the bed. Control counts were made of the number of birds in each cell throughout low tide on two undisturbed days in early and late winter. The response to disturbance was measured from the numbers of birds in the transect during and after the disturbance. As the observer walked to the first cell, birds took flight or walked away so that none were found within a certain distance. Although this local exclusion distance was only measured accurately within the transect, estimates of the distances to nearest birds in other directions suggested that this measure was comparable to that surrounding the observer. Birds continued to move away from the observer for about 15 min after which the exclusion distance remained relatively stable, even if the observer remained in place for two hours (Fig. lb). The exclusion distance was highly variable because it could change by 25 m if a single bird moved between two cells, and so was not used in the analyses. Instead, the local response distance was mea- sured, approximately 3 min after the start of disturbance (i.e. when the exclusion distance had stabilized), as the distance from the observer to the nearest cell in which the number of birds was unchanged from the mean number present before the disturbance. The local return time after the observer left the bed was measured as the time taken for the number of birds in the nearest cell to return to the number observed before the disturbance. Statistical analysis The following variables were related to the three measures of response to disturbance: feeding effort = mean proportion of birds feeding on the bed from the time it became fully exposed to the disturbance (mean =.82, range =.5-.95); number of birds on bed =mean number of birds on the bed during the same period (mean = 313, range = 21-415); stage of season = number of days since 1 August (mean = 12, range = 8-217); temperature = mean of maximum and minimum temperatures (oc) recorded in the 24 hours before 17. on the day of the experiment (mean = 1.7, range = 36 JOURNAL OF AVIAN BIOLOGY 33:4 (22)

53613 Fig. 2. Numbers (a, c) and feeding effort (b, d) of oystercatchers on the whole mussel bed (a, b) and in each transect cell (c, d) during two undisturbed days in early (16 Nov 94) and late (3 Jan 95) winter. In (c, d) the error bars are maximum and minimum values recorded in each cell while the bed was fully exposed. (a) Bed-wide numbers 4 4 O e. 3 - o o o 1. (b) Bed-wide feeding effort S.75 o C o o.- 9 It %-.5 S2 o E oo o e Late winter z -C.25 o Early winter. 1 2 3 1 2 3 Time since first exposure (min) Time since first exposure (min) (c) Local numbers 1 1. I (d) Local feeding effort E 4- z ".25 2 a.. 5 1 15 5 1 15 Distance along transect (m) Distance along transect (m) 2.-23.); day length = hours of daylight on day of experiment (mean = 11.6, range = 9.3-16.7). Feeding effort and the number of birds on the bed were measured from a number of counts, rather than from the last count before the disturbance, in order to account for short-term fluctuations after minor disturbances. Stage of season is related to the general decline in feeding conditions during winter, while the other variables measure specific aspects of the birds' energy demands or feeding conditions: temperature = thermoregulatory energy demands; day length = time available for feeding during the hours of daylight; number of birds on bed = strength of interference. In the analyses we separately related stage of season and feeding effort to each response to disturbance (see below). By performing two tests, we increased the chances of obtaining spurious significant results, and so performed a Bonferroni correction (Zar 1984) to adjust the significance level in these tests from.5 to.25. Results Behaviour in the absence of disturbance The bed-wide control counts showed that, in the absence of disturbance, feeding birds used the bed throughout the low tide period and their numbers were relatively stable between the bed being fully exposed by the receding tide, approximately 1 min after first exposure, and the advancing tide starting to cover the bed (Fig. 2a, b). The local control counts showed that all cells of the experimental transect were occupied continuously by feeding birds while the bed was fully exposed (Fig. 2c, d). Neither feeding effort (one-way ANOVA; early winter, F = 2.32, df= 42,5, P =.6; late winter, F = 1.43, df=42,5, P=.233) nor bird numbers in late winter (one-way ANOVA; F = 1.49, df = 42,5, P =.215) varied significantly among the different transect cells. In contrast, bird numbers in early winter did vary among cells (one-way ANOVA; F = 3.73, df= 42,5, P =.7), possibly due to local variation in feeding conditions, with bird numbers between 75 and 125 m from the start of the transect being lower than in the rest of the transect (Fig. 2c). However, the method of measuring the local response distance accounted for this variation as it compared the numbers of birds present before and after disturbance. Therefore, we did not consider that this variation would affect the experimental results. As the control counts showed that both the entire bed and the transect were used throughout the low tide exposure period, any absence of birds during experiments could be attributed to the disturbances themselves rather than any other factors. Response to disturbance The bed-wide disturbance always caused all birds on the bed to take flight, and the local disturbance always excluded birds from an approximately circular area around the observer. There was considerable variation JOURNAL OF AVIAN BIOLOGY 33:4 (22) 361

53614 in the three measures of the response to disturbance: the bed-wide return time ranged from 5 to 32 min bird-1 (mean = 18, sd = 1, n = 15), the local response distance from 9 to 14 m (mean = 123, sd = 19, n = 27) and the local return time from 5 to 6 min (mean = 27, sd = 19, n = 15). Seasonal changes in response to disturbance Variation in the response to disturbance was related to the stage of the season (Fig. 3). Later in winter, birds returned more rapidly after the bed-wide disturbance (bed-wide return time = 29.4 -.934 stage of season; P =.5, r2 = 46.%), approached the local disturbance more closely (local response distance = 138 -.13 stage of season; P=.18, r2= 2.4%) and returned more rapidly after the local disturbance (local return time = 48.7 -.195 stage of season; P =.4, r2 = 48.7%). All results were significant at the threshold P value of.25 derived from the Bonferroni correction. Seasonal changes in feeding effort The study aim was to determine whether the response to disturbance was related to changes in feeding conditions, and hence the difficulty birds were having in meeting their requirements. The explanatory variables each measured separate aspects of the birds' feeding conditions (see above), and so we investigated, using step-wise multiple regression, whether any combinations of stage of the season, temperature, day length and number of birds on the bed were related to the response to disturbance. However, no single variable consistently explained a higher proportion of the variance in the response to disturbance than stage of the season, and no combinations of variables were selected. The final model for bed-wide return time only included stage of the season, that for local response distance only the number of birds on the bed, and that for local return time only temperature (P <.1 in all cases). An explanation for the lack of significant combined relationships is that each variable was correlated with stage of the season (Pearson correlation coefficients with stage of season; temperature = -.852, P <.1; day length= -.675, P <.1; number on bed =.353, P =.19). To overcome this problem, we used the proportion of birds feeding on the bed prior to the disturbance as a single explanatory variable, expected to provide an index of the difficulty oystercatchers were having in meeting their requirements. As a test of this, the proportion of birds feeding was compared in a stepwise multiple regression (using data collected during experimental and control days) with stage of the season, temperature, the number of birds on the bed and day length. As it was a proportion, a logistic transformation (logit = In(feeding effort/(1 - feeding effort))) was applied to the proportion of feeding birds before the (a) Bed-wide return time 4-3 - * E 2 -.) 1-5 1 15 2 Days since 1st August (b) Local response distance 15-125 - 1 - r 75 5-25 5 1 15 2 6 Days since 1st August (c) Local return time E 4- E 2 - * * * O 5 1 15 2 Days since 1st August Fig. 3. Relationships between stage of the season and the responses of oystercatchers to disturbance: (a) return time after bed-wide disturbance; (b) response distance during local disturbance; and (c) return time after local disturbance. The lines show relationships fitted using linear regression (see text for coefficients). 362 JOURNAL OF AVIAN BIOLOGY 33:4 (22)

53615 1. c" O (.8 o o a S o o *..6 o oo * Temperature lower than expected o Temperature higher than expected I,.4 I I 5 1 15 2 Days since 1 August Fig. 4. Relationship between oystercatcher feeding effort and stage of the season and temperature. The open circles show days on which temperature was higher than expected and the closed circles days on which temperature was lower than expected. Expected temperatures were calculated by regressing temperature against stage of the season (temperature = 18.7-.67 stage of season; P <.1, r2 = 72.5%). feeding effort; P =.2, r2 = 53.1%). All results were significant at the threshold P value of.25 derived from the Bonferroni correction. Oystercatchers returned more quickly after disturbances when they were having more difficulty meeting their energy demands, 4 3- E 2 1 (a) Bed-wide return time.5.6.7.8.9 1. Proportion feeding before disturbance analysis. The regression selected stage of the season and temperature as explanatory variables, showing that the proportion feeding increased later in the season, but was particularly high on unusually cold days (logit = 1.76 +.596 stage of season (P =.1) -.691 temperature (P =.2); r2 = 8.2%; n = 45; Fig. 4). An explanation of why feeding effort was related to a combination of variables, despite the correlation between them, is that, by combining data from all experiments and controls, a larger sample size (n = 45) was possible than for tests on the separate responses to disturbance (n = 15, 27 and 15 for bed-wide return time, local response distance and local return time respectively). Therefore, the proportion of birds feeding did provide an index of the difficulty birds were having meeting their energy demands, and thus the potential starvation risk of avoiding disturbances. Feeding effort and response to disturbance In order to test whether variation in the response to disturbance was related to the difficulty birds were having meeting their requirements, bed-wide return time, local response distance and local return time were regressed against the proportion of birds feeding on the bed prior to the disturbance (Fig. 5). When they were spending more time feeding, birds returned more rapidly after the bed-wide (bed-wide return time = 61.7-54. feeding effort; P =.2, r2 = 53.%), approached the local disturbance more closely (local response distance = 183-72.6 feeding effort; P =.2, r2 = 19.9%) and returned more rapidly after the local disturbance (local return time = 14.3-133.5 15 125 a 1 a) c 75? 5 25 (b) Local response distance.5.6.7.8.9 1. Proportion feeding before disturbance 6 - E 4- a) E 2 (c) Local return time I.5.6.7.8.9 1. Proportion feeding before disturbance Fig. 5. Relationships between oystercatcher feeding effort and their response to human disturbance: (a) return time after bed-wide disturbance; (b) response distance during local disturbance; and (c) return time after local disturbance. The lines show relationships fitted using linear regression (see text for coefficients). JOURNAL OF AVIAN BIOLOGY 33:4 (22) 363

53616 and hence the starvation risk of avoiding the disturbance was greater. Discussion Oystercatchers responded less to disturbances later in winter when they needed to spend longer feeding to meet their requirements. They may have responded less either because the starvation risk of avoiding disturbance was greater or because their perceived predation risk of the disturbance was lower. The starvation risk of lost feeding time and increased energy cost of flying away from a disturbance did increase during winter. Later in winter, birds' energy requirements increase as the temperature falls, and the feeding conditions become poorer as food quality declines and interference competition intensifies (Stillman et al. 1996). Oystercatcher mortality due to starvation occurs mostly in late winter (Goss-Custard et al. 1996, Stillman et al. 2). As they need to spend longer feeding to survive, oystercatchers have less spare time in which to compensate for disturbance late in winter. The perceived risk of the disturbances could have changed if the type of disturbance changed or if birds became habituated to the disturbances. Disturbances were the same in all replicates, eliminating the first possibility. Habituation to the disturbance itself is unlikely to have been the major cause because experiments were conducted so infrequently, once about every 14 days, and experiments were designed to mimic the types of human behaviour that birds would experience anyway. Fur- thermore, individual oystercatchers return winter after winter to the Exe estuary (Durell et al. 2) which probably allows them to habituate in the long-term to disturbance from people. However, we cannot eliminate the possibility that part of the seasonal trend in the response to disturbance was due to habituation to general human activities on the estuary. Whatever the precise cause, the experiments still showed that oystercatchers responded less when they were most vulnerable. Experiments were only conducted on spring tides because the mussel bed did not fully expose on neap tides. Oystercatchers have more difficulty meeting their energy requirements on neap tides because a smaller area of mussel bed is exposed and the upper-shore mussels which are exposed are of a lower quality than the lower-shore mussels exposed on spring tides (Goss- Custard et al. 1993). This means that interference competition is stronger and that birds must consume more mussels to meet their daily requirements. We are confident that a seasonal trend in the response to disturbance will also occur during neap tides. However, as birds have more difficulty meeting their requirements during neap tides, it is possible that behavioural re- sponses to disturbance will be lower on neap tides than those on spring tides. The experiments were conducted on a mussel bed on which the background level of disturbance was relatively low, so that the controlled disturbance was usually the only major source of disturbance during the low tide period. Different magnitudes of response would have been recorded had the experiments been conducted at another site which had a different background level of disturbance. For example, Urfi et al. (1996) showed that oystercatchers flushed at greater distances in parts of the Exe estuary in which people were encountered less frequently, and Smit and Visser (1993) showed similar effects for a range of waders in the Wadden Sea. Such a relationship should not always be expected, however; if contact with humans is detrimental, birds may show stronger responses in areas were they contact humans more often (Smit and Visser 1993). After being disturbed most birds flew to and settled on the sand ridge near the bed, which involved a flight of less than about 5 m. Given the small amount of variation in flight distance, and the probably low energetic cost of such short flights, changes in the starva- tion risk of disturbance probably arose mainly from the amount of lost feeding time. The risk of this depends on the ability of birds to compensate afterwards by either feeding for longer or increasing their intake rate. Although oystercatchers feeding on cockles Cerastoderma edule may increase their intake rate when the time available for feeding is reduced substantially (Swennen et al. 1989), mussel-feeding oystercatchers on the Exe have not been shown to increase their intake rate after disturbance (Urfi et al. 1996). Furthermore, they do not increase their rate of mussel consumption to compensate for an overwinter decline in the flesh content of individual mussels of almost 5%, even though the decline in prey quality contributes impor- tantly to the starvation of birds (Goss-Custard et al. 21). It is likely, therefore, that compensation on the Exe occurs by feeding for longer, and so the birds' ability to do so depends on their amount of spare time. The experiments measured the short-term displacement of birds caused by disturbance, but in common with virtually all other disturbance studies, did not measure the survival consequences. To do this, the results need to be incorporated into a model which predicts the consequences of displacement for the individuals concerned and those feeding in areas to which displaced birds move. The important factors are the amount of time lost and extra energy expended during the disturbance, the amount of spare time available in which to compensate, and the increased strength of interference and depletion caused by the increased bird density in non-disturbed areas. The influence of disturbance on mortality in this system has been investigated using these experimental results and a behaviour-based 364 JOURNAL OF AVIAN BIOLOGY 33:4 (22)

53617 model (Goss-Custard et al. 2, Stillman et al. 21, West et al. 22). An assumption often made in disturbance studies is that disturbance is more serious (e.g. causes more animals to die) when the behavioural response to disturbance is greater (Gill et al. 21). Our results show that the opposite may be true. Oystercatchers responded more to disturbance at times when they were more able to compensate for the lost feeding time associated with this response. We suggest that in order to overcome this problem, studies should measure both the behavioural response to disturbance and the difficulty animals are having meeting their requirements (e.g. the proportion of time spent feeding). Conservation effort should be directed towards species which show a large behavioural response to disturbances, even though they are already having difficulty meeting their requirements. Acknowledgements - We are very grateful to Richard Caldow, Selwyn McGrorty and Andy West for many useful discussions, and Sandra Satterly and Alan Child for supplying the weather data. This work was funded by the Directorate-General for Fisheries of the European Union. References Davidson, N. and Rothwell, P. 1993. Introduction. - Wader Study Group Bulletin 68, Special Issue: 1-2. Durell, S. E. A. le V. dit, Goss-Custard, J. D., Clarke, R. T. and McGrorty, S. 2. Density-dependent mortality in oystercatchers Haematopus ostralegus. - Ibis 142: 132-138. Gill, J. A., Sutherland, W. J. and Watkinson, A. R. 1996. A method to quantify the effects of human disturbance on animal populations. - J. Appl. Ecol. 33: 786-792. Gill, J. A., Norris, K. and Sutherland, W. J. 21. Why behavioural responses may not reflect the population consequences of human disturbance. - Biol. Cons. 97: 265-268. Goss-Custard, J. D. and Verboven, N. 1993. Disturbance and feeding shorebirds on the Exe estuary. - Wader Study Group Bulletin 68, Special Issue: 59-66. Goss-Custard, J. D., Durell, S. E. A. le V. dit, McGrorty, S. and Reading, C. J. 1982. Use of mussel, Mytilus edulis, beds by oystercatchers, Haematopus ostralegus, according to age and population size. - J. Anim. Ecol. 51: 543-554. Goss-Custard, J. D., Durell, S. E. A. le V. dit, Goater, C. P., Hulscher, J. B., Lambeck, R. H. D., Meininger, P. L. and Urfi, J. 1996. How oystercatchersurvive the winter. - In: Goss-Custard, J. D. (ed.). The oystercatcher: from individuals to populations. Oxford University Press, Oxford, pp. 133-154. Goss-Custard, J. D., Stillman, R. A., West, A. D., McGrorty, S., Durell, S. E. A. le V. dit and Caldow, R. W. G. 2. Role of behavioural models in predicting the impact of harvesting on populations. - In: Gosling, M. and Suther- land, W. J. (eds). Behaviour and Conservation. Cambridge University Press, Cambridge, pp. 65-82. Goss-Custard, J. D., West, A. D. and Durell, S. E. A. le V. dit 1993. The availability and quality of the mussel prey (Mytilus edulis) of oystercatchers (Haematopus ostralegus). - Neth. J. Sea Res. 31: 419-439. Goss-Custard, J. D., West, A. D., Stillman, R. A., Durell, S. E. A. le V. dit, Caldow, R. W. G., McGrorty, S. and Nagarajan, R. 21. Density-dependent starvation in a vertebrate without significant depletion. - J. Anim. Ecol. 7: 955-965. Hill, D., Hockin, D., Price, D., Tucker, G., Morris, R. and Treweek, J. 1997. Bird disturbance: improving the quality and utility of disturbance research. - J. Appl. Ecol. 34: 275-288. Hockin, D., Ounsted, M., Gorman, M., Hill, D., Keller, V. and Barker, M. 1992. Examination of the effects of disturbance on birds with reference to the role of environmental impact assessments. - J. Environ. Manage. 36: 253-286. Kersten, M. and Piersma, T. 1987. High levels of energy expenditure in shorebirds; metabolic adaptations to an energetically expensive way of life. - Ardea 75: 175-187. Smit, J. C. and Visser, G. J. M. 1993. Effects of disturbance on shorebirds: a summary of existing knowledge from the Dutch Wadden Sea and Delta area. - Wader Study Group Bulletin 68, Special Issue: 6-19. Stillman, R. A., Goss-Custard, J. D., Clarke, R. T. and Durell, S. E. A. le V. dit 1996. Shape of the interference function in a foraging vertebrate. - J. Anim. Ecol. 65: 813-824. Stillman, R. A., Goss-Custard, J. D., West, A. D., Durell, S. E. A. le V. dit, Caldow, R. W. G., McGrorty, S. and Clarke, R. T. 2. Predicting mortality in novel environments: tests and sensitivity of a behaviour-based model. - J. Appl. Ecol. 37: 564-588. Stillman, R. A., Goss-Custard, J. D., West, A. D., McGrorty, S., Caldow, R. W. G., Durell, S. E. A. le V. dit, Norris, K. J., Johnstone, I. G., Ens, B. J., van der Meer, J. and Triplet, P. 21. Predicting oystercatcher mortality and population size under different regimes of shellfishery management. - J. Appl. Ecol. 38: 857-868. Sutherland, W. J. 1996. From individual behaviour to population ecology. - Oxford University Press, Oxford. Swennen, C., Leopold, M. F. and de Bruijn, L. L. M. 1989. Time-stressed oystercatchers, Haematopus ostralegus, can increase their intake rate. - Anim. Behav. 38: 8-22. Urfi, A. J., Goss-Custard, J. D. and Durell, S. E. A. le V. dit 1996. The ability of oystercatchers Haematopus ostralegus to compensate for lost feeding time: field studies on individually marked birds. - J. Appl. Ecol. 33: 873-883. West, A. D., Goss-Custard, J. D., Stillman, R. A., Caldow, R. W. G., Durell, S. E. A le V. dit and McGrorty, S. 22. Predicting the impacts of disturbance on wintering waders using a behaviour based individuals model. - Biol. Cons. 16: 319-328. Wiersma, P. and Piersma, T. 1994. Effects of microhabitat, flocking, climate and migratory goal on energy expenditure in the annual cycle of red knots. - Condor 96: 257-279. Zar, J. H. 1984. Biostatistical Analysis. - Prentice-Hall, London. (Received 6 August 21, revised 21 January 22, accepted 4 February 22.) JOURNAL OF AVIAN BIOLOGY 33:4 (22) 365