Site selection and resource depletion in black-tailed godwits Limosa l. limosa eating rice during northward migration


  • Pedro M. Lourenço,

    Corresponding author
    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750AA Haren, The Netherlands
    Search for more papers by this author
  • Freek S. Mandema,

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750AA Haren, The Netherlands
    2. Centro de Biologia Ambiental, Museu Nacional de História Natural, Rua da Escola Politécnica 58, 1250-102 Lisboa, Portugal
    Search for more papers by this author
    • Present address: Community and Conservation Ecology Group (COCON), Center for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750AA Haren, The Netherlands

  • Jos C.E.W. Hooijmeijer,

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750AA Haren, The Netherlands
    Search for more papers by this author
  • José P. Granadeiro,

    1. Centro de Biologia Ambiental, Museu Nacional de História Natural, Rua da Escola Politécnica 58, 1250-102 Lisboa, Portugal
    Search for more papers by this author
  • Theunis Piersma

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750AA Haren, The Netherlands
    2. Department of Marine Ecology, Royal Netherlands Institute for Sea Research (NIOZ), PO Box 59, 1790 AB Den Burg, Texel, The Netherlands
    Search for more papers by this author

Correspondence author. E-mail:


1. During migratory stopovers, animals are under strong time stress and need to maximize intake rates. We examine how foragers react to resource depletion by studying the foraging ecology and foraging site selection of black-tailed godwits Limosa l. limosa staging in rice fields during their northward migration stopover (January–March 2007).

2. We analysed godwit abundance and foraging behaviour, sampled the availability of rice in the fields and used the functional response model to predict the giving-up density (GUD) of rice kernels when godwits should give up a rice field. Sightings of individually colour-marked birds were used to verify whether individuals moving between rice fields confirmed the predicted GUD.

3. Black-tailed godwit intake rates at different rice densities fitted Holling’s functional response curve. The predicted GUD of rice necessary to balance allometric estimates of daily energy expenditure (DEE) and measured time budgets were confirmed by GUD measured in the field.

4. Individually marked birds moved towards rice fields with higher rather than lower rice densities more often than randomly expected. These birds increased the measured intake rates after this move.

5. Godwit foraging caused a decrease in the rice density of individual fields during the stopover period. Despite this, overall intake rates remained constant as godwits reacted to resource depletion by moving to a new foraging site as soon as their intake rate falls below the required levels to achieve DEE.


During migratory stopovers, animals gather in large numbers in relatively confined areas, needing to quickly find enough food to replenish their energy stores to continue their migratory journeys (Newton 2006). Understanding staging strategies reveals much about migratory systems (Lindström 1995). Interpreting how food availability shapes habitat use can provide great insights to what foraging decisions are behind the refined way in which migrants make the best of these critical stop-over periods.

To correctly assess the quality and availability of foraging habitat, an understanding of the processes underlying their foraging activity and resource use is necessary (Goss-Custard et al. 2003; van Gils et al. 2004). The carrying capacity of a given habitat unit primarily depends on food abundance and the degree to which losses are replenished. In the simplest approach, intake rates depend on the density of available food resources and the forager’s functional response to these resources (Holling 1959; Piersma et al. 1995; van Gils et al. 2004). Functional response relationships predict an upper limit to intake rates, because of the necessary handling time spent on each food item. It also predicts a decrease in intake rates at lower food densities, because a forager has a limited instantaneous area of discovery, thus spending more time to find each food item when density decreases. This relationship is described by the following equation (for more details see Holling 1959; Piersma et al. 1995):


where N is the number of prey ingested, a is the instantaneous area of discovery (m2 s−1), D is the food density (m−2), T is the total time spent foraging (s), and Th is the handling time (s). A patch should be abandoned when the local intake rate falls below the required level to balance the current energy budget (Stephens & Krebs 1986; Piersma et al. 1995; van Gils et al. 2004).

Ideal and free foragers would distribute themselves proportionally to the food abundance of each patch in what is known as an Ideal Free Distribution (Fretwell & Lucas 1970; Sutherland 1996). However, foragers frequently are not free and not ideal, as movements are never cost-free and the identification of relative resource availabilities will be constrained at some level (van Gils et al. 2006b). Also, the metabolic cost of foraging might differ between patches, and different patches may vary in terms of predation risk, both potentially causing ‘unexpected’ situations where the individual fitness may be maximized by foraging in the patches with lower food abundance (van Gils et al. 2004).

Foraging on non-renewable resources will eventually deplete food stocks, causing a decrease in the quality of a particular habitat unit (Vickery et al. 1995; Gill, Sutherland & Norris 2001b; van Gils et al. 2006b). Furthermore, when changes in the quality of the habitats occur, animals are faced with the decision to either stay at a site or to attempt to find a better one. The decision to move is believed to be based on the density of prey; the prey densities measured in the field after foragers give up on a unit of habitat are known as giving-up densities (GUD; Brown 1988). Even when the information about the overall availability of food resources in not freely available to the foragers, it is possible that simple patch-leaving rules allow an optimal use of resources (Griffen 2009). Such decisions are particularly important during stopovers, when time is short, information is limited and the price for delays is paid in terms of reduced breeding success and even survival (Newton 2006).

There are several accounts of migrants depleting food resources, both during the winter (e.g. Raffaelli & Milne 1987; Rosa et al. 2008) and at stopover sites (e.g. Schneider & Harrington 1981; Moore & Yong 1991; Nolet & Drent 1998; van Gils & Tijsen 2007). However, this information alone does not imply that resource depletion has a negative effect on staging migrants. Foraging animals may increase foraging effort or simply move to better foraging locations to maintain required levels of food intake despite food depletion or competition (van Gils et al. 2006b; Lewis, Esler & Boyd 2008). Digestively constrained foragers can also respond by increasing the processing capacity of the digestive system (van Gils et al. 2006a).

The black-tailed godwit Limosa limosa limosa (L.) is a long-lived migratory bird, breeding in northern Europe and wintering in Western Africa. During northward migration, large numbers of these birds stopover in Iberia, where they forage in large flocks that gather in rice field areas (Kuijper et al. 2006; Lourenço & Piersma 2008). This concentration of animals in a relatively small number of sites, and the fact that they forage almost exclusively on spilled rice kernels (Lourenço & Piersma 2008), enabled us to measure and correlate resource depletion, foraging activity and site selection of staging migratory birds. This study aims to: (i) determine the extent to which black-tailed godwits deplete their food source during their stopover in Iberia; (ii) examine how resource depletion affects intake rates and the use of different foraging sites; (iii) analyse whether birds move to better sites in response to resource depletion; and (iv) understand which foraging decisions could explain the observed site selection.

Materials and methods

Field work took place in the winter 2006–2007, in a number of rice field areas covering 2547 ha around the estuaries of the Tejo and Sado rivers on the Portuguese west coast (for more details see Lourenço & Piersma 2008). All studied rice plantations were surveyed at least once per week during the period of stopover between late December and early March (Kuijper et al. 2006). Only 11 rice fields, covering 28 ha, were used by the godwits and are part of our analysis (Fig. 1). Whenever a godwit flock was found, the number of birds was counted and rice abundance was measured.

Figure 1.

 Location of the study areas (in black) around the estuaries of the Tejo and Sado rivers, in central Portugal..

At each rice field, 15 birds were used for behavioural observations each week. These were randomly selected by using only every 100th scanned bird for focal observation. Each individual was observed for a maximum period of 120 s (birds observed for <60 s were excluded from further analysis) and the number of probes and swallows were counted. These were later converted into rates per minute. Probes were defined as periods when the bill was inserted in the sediment, a new probe only started after the bill tip had been out of the sediment. Food intake rates were defined as the number of swallowing movements following upward movements of the bill. Recent work in Spain showed that godwits always take only one rice seed at a time (J.A. Masero, pers. comm.; Santiago-Quesada et al. 2009), so we assume that each swallowing movement represents the ingestion of a single rice kernels. The proportion of each behaviour in a flock was scored in three categories (foraging, roosting and preening) in samples of 100 birds. In the last 2 weeks of the study (late February–early March), some flocks only had roosting and preening birds. These were excluded from the analysis as they probably represented birds that were ready to migrate; such birds may not forage because they are undergoing body composition changes in preparation for the long-migratory flight (Piersma 1998; Landys-Ciannelli, Piersma & Jukema 2003).

Rice abundance was measured in each rice field by collecting four samples at random locations using a flat shovel that was able to cut through the hard roots in the soil. At each field, samples were collected once per week since the first time a godwit flock was detected until 1 week after the godwits stopped using that field. Four fields were only sampled twice as they were used for a single week (fields 1, 2, 3 and 4) and the remaining fields were sampled 3–6 times. In three cases (fields 7, 9 and 10) we were unable to sample every week and so the sampling frequency is more sparse. Each sample was a 10 × 10 cm2, with a depth of roughly 12 cm, to correspond with the maximum bill size of an adult black-tailed godwit (Schroeder et al. 2008). The coefficient of variation between each set of four samples averaged 0·56 ± 0·27, so we believe this sampling effort was sufficient to precisely characterize local rice densities in each occasion. One field had extremely high rice densities, as it was not harvested because of a flood. This field was excluded from the rice depletion analysis (see below), as the densities were 100× larger than anywhere else and were not comparable to the other fields. However, intake rates in this field were used to estimate plateau levels of intake rate.

We fitted the measured intake rates to Holling’s functional response equation. For this purpose we used the intake rates from the field with extremely high rice densities to calculate the handling time (Th). The probing success rate was maximum at this site, achieving 60%, and the high rice densities suggest this field birds needed nearly no time to search, thus Th was simply the reciprocal of the intake rate. As we had no empirical measure of the instantaneous area of discovery (a), we calculated the value that maximized the fit to our data, using the least squares method. Afterwards, we calculated the minimum necessary intake rate for godwits to compensate their daily energy expenditure (DEE; calculated using the allometric equation in Kersten & Piersma 1987), assuming only diurnal foraging at 9·5 h day−1 (average day length in Portugal during the study period), an energy density of 0·35 kJ per rice kernel, and a digestive efficiency of 60% (L. Zwarts, pers. comm.). Finally, we extrapolated the expected rice GUD from this minimum intake rate (Piersma et al. 1995).

To test whether godwit foraging had an effect on rice densities, we fitted hierarchical linear models to our observations. We used a cumulative measure of godwit use of the rice fields by calculating cumulative godwit days as the average number of godwits counted on a rice field multiplied by the number of days between the first and last time godwits were detected in that particular field. On a few occasions there were isolated days when a field was found empty, but used again in the next day, so we only considered the godwits stopped using a field after two or more days of absence. These isolated days with zero godwits were also included when calculating the average number of godwits and the cumulative godwit days. Furthermore, as the presence of straw in the samples was found to positively affect the amount of rice (samples with straw, 4468 ± 1010 rice kernel m−2; samples without straw, 365 ± 530 rice kernels m−2; t = 8·4, d.f. = 207, < 0·001), this was corrected by including a ‘presence of straw’ factor in the analysis, as this factor may affect the distribution of rice densities within a field and influence the ability of godwits to find rice kernels. Each sample in a rice field was nested within a factor for field. To include time in the analysis, observations were assigned to weeks and a factor ‘week’ was used. All analyses were made using the ML Win software (ML WIN, Centre for Multilevel Modelling, University of Bristol, Bristol, UK), which uses Bayesian models by Markov Chain Monte Carlo method, as well as iterative bootstrapping to estimate the generalized multilevel linear models (Goldstein 1995). To compare different models, a test based on likelihood ratios was used, which can be approached by a chi-square distribution.

To confirm that godwits move towards rice fields with higher rice abundances we analysed sightings of individually marked birds. In recent years, many black-tailed godwits have received individual colour ring combinations for individual identification in the field (e.g. Gunnarsson et al. 2005; van den Brink et al. 2008). A total of 172 individuals with colour rings were detected during our work. Both European subspecies of godwits, Limosa l. limosa and L. l. islandica, occur in Portugal during the winter, the first during northward migration stopover, the second as a winterer. Despite being mostly habitat segregated, with limosa in freshwater habitats and islandica in brackish and salt water habitats (Kuijper et al. 2006), there is some overlap in the rice fields (P.M. Lourenço pers. obs.). To ensure that only staging birds were analysed, we analysed movements of birds ringed in The Netherlands during the breeding season, which can be unequivocally assigned to the nominate subspecies. This left us with 84 individuals to work with, of which for only 31 we observed movement between rice fields (only one movement per bird). The weekly re-sighting rate of marked birds was 32% (Lourenço et al. in press), so an important proportion was missed every week; however, the fact that all observed movements were one way only, gives some confidence that these reflect actual movements. For each of these 31 movements we determined whether the move occurred towards a field with higher rice density (at least 100 rice kernels m−2 more), similar rice density (give or take 100 rice kernels m−2) or lower rice density (at least 100 rice kernels m−2 less). The movements were compared with the proportional availability of rice fields with higher, similar or lower rice densities at the date of the movement. We used all sampled fields for this analysis, even the ones located at large distances from the origin of the movement, as we have evidence that movements often occur at both short scales and large scales (Lourenço & Alves in press). As proportions of field types always sum to 1 and are not inter-independent (unit-sum constraint; see Aitchison 1986), we used compositional analysis to examine our data (for more details see Lourenço & Piersma 2008). Throughout the text, means are presented ±SE.


The proportion of birds foraging was estimated at 89·4 ± 7·2% (= 86 flocks) which can be assumed to be the average portion of time spent foraging by each bird (Altmann 1974). The remaining ones were roosting (7·1%) or preening (3·5%). Overall, the probe rate was 15·5 ± 1·3 probes min−1 and the food intake rate was 7·0 ± 1·2 food items min−1 (in both cases, = 270). Probing success varied between 37% and 60%. None of these variables varied significantly along the stopover period, nor had any significant correlation with bird density. Probing success was significantly higher at higher rice densities (r= 0·27, = 132, < 0·001).

In the unharvested field, godwits consumed 14·9 ± 2·7 food items min−1, which translates in an estimated handling time, Th = 4·03 s. We then found the value for a that optimized the fit of our data to Holling’s functional response at 9·4 ± 1·7 cm2 s−1 (r= 0·21, = 132, < 0·01) (Fig. 2). The minimum intake rate to sustain the calculated DEE of 567 kJ day−1 was determined to be 4·73 rice kernels min−1, which gives a GUD of 123 rice kernels m−2.

Figure 2.

 Relationship between average food intake rate and rice abundance. Each data point was coded with a symbol representing the field were the data were collected. Data were fitted to Holling’s functional response equation, assuming Th = 4·03 s. The fit best fit was found with = 9·41 cm2 s−1. The horizontal line shows the minimum intake rate godwits need to achieve their DEE (4·73 kernel min−1) and the vertical line indicates the extrapolated rice giving-up density (122·8 kernels m−2).

We found no significant correlation between initial rice abundance and how intensely a field was used by the godwits (r= 0·18, = 10, > 0·1), indicating a non-proportional distribution of godwits amongst fields of varying food abundance. Despite this, there was a clear negative relationship between the cumulative godwit days and rice abundance (χ= 70·6, d.f. = 1, < 0·001), and rice abundances in each rice field decreased along the course of the season (χ= 12·9, d.f. = 1, < 0·001; Fig. 3). The GUD, measured in the fields after godwits stopped using them, averaged 231 ± 106 kernels m−2 (= 10). However, this average is strongly influenced by the very high final rice density (1150 ± 316 kernels m−2) found in one field (field 4). If we exclude this outlier, we find an average final rice density of 134 ± 49 kernels m−2 (= 9), very similar to the 123 kernels m−2 GUD predicted by the functional response relationship (Fig. 2). This can be interpreted as the rice giving-up densities for godwits foraging on our study areas.

Figure 3.

 Variation in the food abundance in each rice field along the stopover period. Each line represents the linear regression of rice densities on one rice field, and in almost all cases the trend is clearly negative. The dashed line represents the 122·8 kernels m−2 giving-up density (GUD) extrapolated from functional response. On the right side is a zoomed view of the lower part of the rice density scale showing the distribution of observed GUD (excluding field 4). Box represents ±SE, whiskers represent the range.

Our results suggest that godwits move to a different field when local rice densities become too low. Indeed, 77% of the movements of 31 colour-ringed birds were towards fields with higher rice abundance, while only 10% moved into fields estimated to have less rice. Godwits moved to rice fields with higher rice abundance more than randomly expected (compositional analysis, Λ = 0·29, < 0·01), considering the availability of rice fields with different rice densities at the time of each movement (Fig. 4). Furthermore, for 14 of the moving birds we have data on the change in intake rates before and after they moved. The 11 that moved to higher rice densities had an increase in their intake rate (before: 5·6 ± 0·6 food items min−1, after: 7·3 ± 0·5 food items min−1; Wilcoxon matched pairs test: Z = 2·14, = 11, < 0·05). The two birds that moved to lower rice densities had a slight decrease (before: 5·5 food items min−1, after: 5·0 food items min−1), as did the bird that moved to an area of similar rice density (before: 6·1 items min−1, after: 5·3 items min−1).

Figure 4.

 Comparison between the proportion of observed individual movements towards fields with different rice densities (black bars) and expected values based on the availability of rice fields with varying rice densities (grey bars).


The intake rates measured were affected by rice density. With increasing rice densities, the intake rate increased, which is consistent with the idea that foraging is more efficient in habitat units with higher food abundance. Black-tailed godwit food intake rates in the rice fields matched the expectations of Holling’s functional response curve. The fit to Holling’s equation was within the range of values reported by other studies (average r= 0·43 ± 0·24, range: 0·01–0·97, in 23 studies performed on eight different species, Goss-Custard et al. 2006). The instantaneous area of discovery that best fitted our data was 9·4 ± 1·7 cm2 s−1, which is almost three times higher than the calculated value for black-tailed godwits foraging on estuarine bivalves (3·3 ± 0·7 cm2 s−1 in Gill et al. 2001b). This suggests that rice kernels are more easily detected than bivalves, or that the sediment structure of rice fields facilitates the search for small objects (see Piersma et al. 1998). One way in which godwits could increase their efficiency in the rice fields would be if they can detect the presence of straw and associate that to a higher rice density.

The average rice GUD measured in the field closely matched the expected value assuming that godwits would leave a rice field when intake rates no longer sustain the allometrically predicted DEE. There was some variation between fields, but it was not very large, with the exception of field 4 which had considerably higher rice densities after godwits left. This value could be as a result of a sampling error, but could also mean that the foraging costs in that field were much higher (van Gils et al. 2004). In fact, this field was adjacent to a railway bridge, and the frequent passage of trains flushed the godwits. Disturbance is known to significantly reduce habitat quality (Gill, Norris & Sutherland 2001a; van Gils & Tijsen 2007) and could in this case be increasing the costs of foraging, leading to the observed high GUD.

Overall, 120–140 kernels m−2 is likely to be the threshold below which birds react by moving to a new field. This means godwits leave a foraging patch just before the food density drops below the necessary level to maintain their energetic balance. However, in our study the birds are at a migration staging site, where they need to re-fuel for further migration. This means they need to obtain enough energy to survive, plus the necessary energy to replenish their fat stores. Our results suggest that a simple foraging rule, namely, leaving a foraging site when food intake falls below the DEE can be sufficient to guarantee the necessary high intake rates for a staging migratory bird. Such patch-leaving decisions can allow a forager to cope with limitations in their ability to perceive food abundance and optimize habitat use despite those limitations (Griffen 2009).

Ideal Free Distribution models predict that resource patches will be used proportionally to the abundance of food (Fretwell & Lucas 1970). However, we did not find any correlation between rice abundance and godwit use of the different fields. These data suggest that staging black-tailed godwits do not use the rice fields proportionally to food availability. Differential metabolic costs of foraging, predation risk and/or the ability to perceive food abundance can cause foragers to use resource patches in a manner not proportional to food abundances (van Gils et al. 2004; van Gils & Tijsen 2007). Such costs might explain the variation in GUD across fields.

Another factor limiting the ability to perceive food abundance can be unequal food availability within a rice field. In fact, the samples with straw had much higher rice densities. The question is then whether godwits are able to detect these small-scale variations in food distribution. Godwits foraging in mudflats seem to use prior knowledge to increase the ability to rapidly perceive the small-scale distribution pattern of their food (Dias, Granadeiro & Palmeirim 2009). It is likely that at such small scales this can also happen in the rice fields, for instance by focusing the foraging effort on patches with straw. However, if this or other small-scale factors affecting the availability of food and are not perceived by the godwits, this would cause resource depletion to have stronger effects on intake rates.

One further limitation is likely to affect how the godwits perceive food availability. The rice fields used by the godwits may be as much as 70 km apart in the meandering valleys of the Tejo and Sado tributaries. How are they able to detect the best foraging sites at this spatial scale? Previous knowledge is not likely of much help, as the quality of the rice fields varies much between years (P.M. Lourenço pers. obs.). However, godwits prefer fields that have been ploughed and that are flooded (Lourenço & Piersma 2008), and these characteristics are easy to see from the air.

Black-tailed godwits frequently use the rice fields at very high densities, with flocks of over 10 000 individuals using a single field of no more than a few hectares. Such strong flocking behaviour suggests the need to defend against predators or possibly an important role of social information to find the best foraging sites (Németh & Moore 2007). The raptor community in our study area mostly consists of common buzzard Buteo buteo (L.), marsh harrier Circus aeroginosus (L.) and common kestrel Falco tinunculus L. (Lourenço in press), all unlikely to represent a significant threat to adult godwits. Nevertheless, they do cause an alarm reaction in godwit flocks (P.M. Lourenço pers. obs.), and so might be influential in the choice of foraging sites. The raptor densities were higher around the Tejo than in the Sado estuary (Lourenço in press). That most rice fields used by the godwits in 2007 were in the Sado estuary is consistent with this hypothesis. On the other hand, flock sizes are on average higher in the Tejo, and rice GUD in the Tejo (118 ± 33 kernels m−2, = 2) were similar to the ones in the Sado (excluding field 4, 142 ± 44 kernels m−2, = 7), so predation costs do not seem to be of importance for deciding when to abandon a foraging patch. In migrating birds the marginal value of energy is rather high, as they need to fuel-up as fast as possible, which leads to low predation costs (Brown & Kotler 2004). Indeed, transient bar-tailed godwits Limosa lapponica (L.) in the Wadden Sea do not seem to avoid dangerous places either (Duijns et al. 2009).

As in the rice fields used by godwits there are no other important rice consumers (Lourenço & Piersma 2008), our results clearly show that godwits flocks have a strong effect on rice densities. On average, they reduced rice availability by 50 ± 20% (= 10). Such a decrease is comparable to recorded levels of depletion of insect larvae by passerines at a stopover site in the Gulf of Mexico (30–65% in Moore & Yong 1991), of mudflat invertebrates by staging shorebirds at a coastal site in Massachusetts (with one exception, 30–90% in Schneider & Harrington 1981), and of chironomid larvae by shorebirds at an inland river mudflat in Hungary (87% in Székely & Bamberger 1992). Despite this reduction in rice abundances in individual fields, and the observed functional response of food intake rates, the overall average intake rate did not vary along the stopover period. Again, this seems to confirm the idea that godwits abandon fields when the local rice densities depress their intake, to search for fields with higher rice densities, similar to what is know for Brent geese Branta bernicla (L.) foraging on coastal habitats in the UK (Vickery et al. 1995), red knots Calidris canutus (L.) foraging on mudflats in the Wadden Sea (van Gils et al. 2006b) or Bewick’s swans Cygnus columbianus bewickii (Yarrell, 1830) feeding on harvested sugar beet fields (van Gils & Tijsen 2007).

The presence of individually colour-marked individuals in the foraging flocks provided us with an opportunity to confirm that individual godwits do move from depleted fields to fields with higher rice abundance. Despite the relatively low sample size (= 31 birds), the movements of colour-ringed black-tailed godwits to fields with higher rice abundances much more than expected by chance alone, are consistent with the idea that birds abandon depleted fields to move into areas with higher rice abundance. The fact that the birds that moved to fields with higher rice density had an increase in their intake rates underpins the notion that foraging site selection is driven by food availability.

Clearly, resource depletion is an important factor in the habitat availability and site selection for black-tailed godwits staging on Iberian rice fields. The intensive foraging activity of large godwit flocks depletes the rice kernels in the soil to a point where birds need to move to a new field to find advantageous foraging grounds. As rice kernels in winter are a non-renewable resource, fields that become depleted will no longer represent available habitat for the remaining staging period, thus depletion reduces the availability of good quality foraging sites during the stopover period. This effect can be partially compensated by the phased ploughing of the rice fields (Lourenço & Piersma 2008), which offers new feeding opportunities, but requires frequent switches between sites. Currently, farming practices do provide this favourable situation of fields becoming available along the staging period, but any changes in this status quo could have serious consequences for the quality of this staging habitat for godwits in Iberia.

Migratory birds are often challenged with such unpredictabilities in the spatial and temporal availability of their food (e.g. Davis & Smith 2001). Godwits seem to have evolved the ability to quickly detect the best foraging patches, and by following a simple rule they are able to leave each foraging site before it becomes unprofitable, thus maximizing their overall food intake.


We would like to thank all the landowners who kindly allowed access to their fields and often provided useful information on where to find the flocks. We would also like to thank N. Groen, N. Cidraes-Vieira and all other observers who provided re-sightings of individually colour-marked birds. We also thank J.A. van Gils for very helpful suggestions on early drafts of this manuscript, and T.G. Gunnarsson and two anonymous reviewers for additional comments. This work was funded by the Portuguese ‘Fundação para a Ciência e Tecnologia’ through grant SFRH/BD/21528/2005.