Climate change affects breeding of reed warblers Acrocephalus scirpaceus


  • Lucyna Halupka,

  • Andrzej Dyrcz,

  • Marta Borowiec

L. Halupka (correspondence), A. Dyrcz and M. Borowiec, Department of Avian Ecology, Zoological Institute, Wroclaw University, ul. Sienkiewicza 21, 50-335 Wroclaw, Poland. E-mail:


Between 1970 and 2006 reed warblers Acrocephalus scirpaceus started breeding progressively earlier; both the initiation of breeding (the earliest first egg dates) and peak of breeding (median first egg dates) advanced. Median first egg dates correlated significantly with increasing May–July mean temperatures. However, in contrast to other studies showing the advancement in laying dates, the end of the season did not shift. As a result, the breeding season is now longer increasing re-nesting opportunities. Individuals are able to re-nest 4–5 times, which might have important implications for the species. It was also found that in warmer seasons the population suffered fewer nest losses. Both factors, higher re-nesting potential and a trend toward fewer losses, should lead to increased fitness of individuals in the studied population.

Recent studies have shown that global warming affects numerous species, including birds (Schiegg et al. 2002, Parmesan and Yohe 2003, Møller et al. 2004). Most data come from studies on timing of migration and laying phenology; warmer springtime temperatures were associated with earlier returns from wintering grounds and initiation of breeding (Tryjanowski et al. 2002, Mitrus et al. 2005). There are growing evidence that the impact of global warming varies widely between and within species (Sanz 2003, Visser et al. 2003, 2004). However, the consequences of this change on demographic factors and fitness measurements are poorly understood (Sanz 2002a, Crick 2004, Visser et al. 2004). In addition, the majority of these few detailed studies have been carried out on woodland, hole-nesting species: tits and flycatchers (Visser et al. 1998, Bairlein and Winkel 2000). Most of them have been performed on populations breeding in artificial nest-sites. Hence, there is an urgent need for detailed studies analysing various breeding parameters of species in a wide variety of natural habitats.

In this paper we analyse changes in breeding parameters of reed warblers Acrocephalus scirpaceus, studied during 12 breeding seasons between 1970–2006. The species is a small, long-lived passerine that breeds in the Palaearctic, but spends its winters in Africa. Reed warblers breed in marshland habitats, almost exclusively in reed beds (Schulze-Hagen 1991, Cramp 1992). They have a relatively long breeding season (defined as the time when nests with offspring can be found), which lasts from May until August/September in central and western Europe (Schulze-Hagen 1991, Cramp 1992). Due to nest losses (generally 50% or more), caused mainly by predators, they frequently re-nest and can lay up to 5 clutches per season (Schulze-Hagen 1991, Cramp 1992). Few pairs (no more than 35%) are able to raise two broods successfully during a breeding season (Schulze-Hagen 1991, Cramp 1992, Borowiec 1994). The species forages on and feeds its nestlings different types of food; food items are often acquired outside their territory (Dyrcz 1979, Bibby and Thomas 1985, Schulze-Hagen et al. 1996). Food resources, examined both at our study site and elsewhere, have been found to be rich over the entire season (Bibby and Thomas 1985, Cramp 1992, Dyrcz and Zdunek 1996).


Study period and area

The breeding biology of reed warblers was studied over 12 breeding seasons (1970–73, 1980–83, 1994, 2003, 2005–06) at fishponds in the Stawy Milickie Reserve in southwestern Poland (Dyrcz 1981, Borowiec 1994, Halupka and Wróblewski 1998). Most data were collected on the Sloneczny pond, which is 180 ha in size, with the emergent vegetation covering about 16 ha. Two study plots (respectively 2.5 and 5.5 ha) were established within a belt of reed bed (1km long, 40–160 m wide) that bordered the southern bank of the pond. The emergent vegetation consisted primarily of common reeds Phragmites communis and cattails Typha angustifolia. The banks were overgrown with trees and bushes.

Data collection

Each season, data were gathered from May to August for 5–7 d a week for about 10 h a day. We attempted to locate all reed warbler nests on the study plots. We found a total of 1,241 nests containing eggs or nestlings across the 12 seasons of data collection; between 60 and 201 nests were found each year. Differences in nest numbers across years result mainly from data gathering from one or from both plots. Breeding statistics (e.g. nest losses, clutch size, phenology) did not differ between the two plots (Borowiec 1985, 1994). Nests were typically visited 2–3 times a week, with more frequent visits (every 1–2 d) taking place during egg-laying and before the expected hatching and fledging dates. In 2003, experiments with artificial cuckoo Cuculus canorus eggs, that could potentially bias certain breeding statistics (nest losses and production of fledglings per nest) were conducted (Dyrcz and Halupka 2007). Therefore, for that year, only measurements of laying dates and clutch sizes, obtained before any experimental procedure had started, were used in this paper.

Nest location was determined by observing nest building and feeding of the young by the parents, as well as by systematic searches through the reed bed. Systematic searching was especially employed late in the season, when the reeds were tall and parental behaviour was much less conspicuous. Reed warbler nests are relatively easy to find (Bibby and Thomas 1985, Kleven et al. 2004). Since we found most nests as they were being built, their first egg dates were accurately measured. When a nest with a complete clutch was found, we monitored the nest until hatching day and then backdated to estimate clutch initiation date, assuming one egg laid per d and an incubation period of 11 d. If the nest had nestlings when found, the age of the offspring was determined based on comparisons with known age-associated features (size, colour, degree of eye opening, feather development etc.).

We are reasonably certain we found the earliest nests in each season for the following reasons. First, the arrival of both sexes was successive (Schulze-Hagen 1991, Borowiec 1992) and when the first females arrived, the overall density of males was low, facilitating the identification of pairings (Borowiec 1992). Second, at the beginning of the season, the reeds are short in stature and it is possible to observe pairs even far into the reed bed. Third, after pair formation, male behaviour changes dramatically. Males cease singing and start guarding their mates continuously until laying. Mate-guarding is very conspicuous, since the male remains within 0.5–2 m of its mate most of the time, and lasts a long time (about 7–10 d). Fourth, a pair engaged in nest-building searches for nest material far outside its territory as, at that time of the season, nest material is scarce (Borowiec 1985, Schulze-Hagen 1991). Fifth, our search efforts were extensive. The relatively small study area was searched approximately 6 days a week for about 10 h per day. For these reasons, it would have been difficult to overlook pairing events and/or nests early in the season (cf. Bibby and Thomas 1985).

In 1980–82, 1994, and 2005–06, the population was colour-ringed so it was possible to trace particular individuals throughout the season. We assumed that nestlings surviving through day 10 fledged successfully (Honza et al. 1998). Nest losses were calculated with the Mayfield (1975) method.

Climatic data

In our analyses, we used the mean monthly temperature in April (the month in which birds typically arrived from wintering grounds), and in May, June, July and August (which encompass the breeding season). In most analyses we used mean May–July temperatures, the time period corresponding with laying.

As temperature in the study area was only measured in 1980–1998 (0.4 km from the study plot), we used data obtained from the regional meteorological station in Wroclaw, some 60 km from our site. At both sites, temperature was measured in standard meteorological conditions using similar methods; measurements were taken 4 times a day, every six hours, starting at 6 a.m. Both data sets correlated well (for years in which both were available) when comparing mean May–July temperatures (rs=0.99, n=19, P<0.001), and the median difference between the two sites was minor (0.26° C). Correlations between monthly temperatures were also very strong ( all rs>0.95, n=19 in each analysis, P<0.001).

Data analysis

To detect changes in breeding phenology, we used two measures utilized in similar studies: the earliest first egg dates (e.g Both et al. 2005b, Schaefer et al. 2006), and median first egg dates (e.g. Sanz 2003, Schaefer et al. 2006). The earliest first egg date (the start of laying) was defined as the date of laying of the first egg in the earliest nest in the season. Although in some seasons first pairs began breeding well before the rest of the population, the first egg date of the earliest pair correlated significantly with the median of the 10 earliest pairs in each season (rs=0.874, n=12, P<0.001).

Median first egg date was calculated from all nests found in the season (including first clutches, repeat clutches, and second broods). This measure also correlated significantly with the earliest first egg dates (rs=0.838, n=12, P=0.001). The length of the laying period was defined as the time between the earliest first egg date and the day of laying the last egg in the latest nest in the season (called the end of laying). Fledgling production per nest was the average number of nestlings that survived until the 10th day of life, and hence most likely managed to leave the nest (Honza et al. 1998). Second clutches/broods were defined as those laid after successfully rearing a first brood. Breeding season was defined as the time between arrival from wintering grounds and fledging of the last nestlings.

As the data violated assumptions necessary to conduct parametric analyses (skewed distributions of some variables), nonparametric tests were used (Martin and Bateson 1993, Sokal and Rohlf 1995). All P-values are given for two-tailed tests and are exact (i.e. without normal approximations). We used StatsDirect software. Power estimates for Spearman rank correlation coefficients were calculated using the bootstrap method (5,000 iterations).

Study population

In our breeding population, the first males arrived from the wintering grounds around May 9–11 in the 1980s and around April 15-May 1 from 1990 to 2006 (Borowiec 1992, Witkowski et al. 1995). The first females arrived around May 11–15 in the 1980s and around May 4–9 from 1990 to 2006. About 20% of the birds that bred at the study site in a given year returned the following season (Borowiec 1985, 1994, unpubl. data). Although the density of breeding pairs changed considerably in time and space over the season, the annual maximum breeding density (estimated as the maximum number of pairs breeding simultaneously) remained quite stable across all years (about 14–17 pairs/ha on plot I and 9–11 pairs/ha on plot II). Females laid an average of 2.4 clutches per season. About two-thirds of the females re-nested 1–4 times; the remaining one-third disappeared from the study plot after their first attempt (Borowiec 1985, 1994). As nest losses were high (44–80%), the number of pairs rearing second broods was low (around 10%) in most seasons (Dyrcz 1981, Borowiec 1994). Clutches were incubated for 11 d and the young fledged when they were 10–13 days old (median 12 d).

The potential food resources of reed warblers in the study area were studied in 1982–84 (Dyrcz and Zdunek 1996), from late April (before the birds’ arrival) until late July. The study revealed that the most important prey (Dyrcz 1979) showed seasonal fluctuations; their peaks did not occur at the same time, resulting in high levels of available resources throughout the season.


Between 1970 and 2006, mean breeding season temperatures (April–August) increased significantly. The trend was significant for each month of the season, but the most pronounced rise occurred in April (rs=0.620, n=37, P<0.001). Average temperatures in May–July, the laying period, increased by 2o C (Spearman rank correlation: rs=0.613, n=37, P<0.001, Fig. 1).

Figure 1.

Mean May–July temperature (° C) in 1970–2006.

During the study period, reed warblers started breeding significantly earlier: both the start of laying (the earliest first egg dates) and peak of laying (median first egg dates) advanced. In 2005 and 2006, egg-laying (measured by first egg date of the earliest pair) started 3 weeks earlier than in 1970, on May 10 and 12 vs. May 31, respectively (Table 1, Fig. 2). Likewise, median first egg dates shifted by 18 d, from June 30 in 1973 to June 13 in 2006 (Table 1).

Table 1.  Spearman rank correlation coefficients between the study year and breeding parameters of reed warblers. P-values and bootstrap power estimates (for two tailed tests with α=0.05) are shown. Sample size (number of study seasons) equals 12 for the first four statistics, and 11 for the two last ones.
Breeding parameterrsPPower
Earliest first egg laying date−0.891<0.0010.97
Median first egg date−0.877<0.0010.98
Clutch size0.1190.7000.09
Length of laying period0.6630.0200.67
Nest losses−0.3600.2730.19
Production of fledglings per nest0.2640.4180.11
Figure 2.

The earliest first egg dates of reed warblers in various study years.

Median first egg dates correlated significantly with the increasing mean May–July temperatures (Spearman rank correlation: rs=−0.916, n=12 seasons, P<0.001, Fig. 3). In warmer seasons, the peak of laying advanced. The earliest egg-laying dates correlated significantly with April–May temperatures (rs=−0.619, n=12, P=0.035).

Figure 3.

Relationship between mean May–July temperature (° C) and median first egg dates of the reed warbler.

In contrast, the end of laying did not change significantly over the study period (rs=−0.311, n=12, P=0.320), in most years, last eggs were laid in late July. As a result, the laying period has become significantly longer (Table 1, Fig. 4).

Figure 4.

The length of the laying period in various study years.

Mean clutch size of the population did not change over the study period (Table 1, Fig. 5). Additionally, fledgling production per nest and nest losses did not change significantly between 1970 and 2006 (Table 1). However, we did find that the mean temperature in May–July correlated negatively with the proportion of nests that failed (rs=−0.618, n=11, P=0.048) and there was some evidence of a positive relationship with the number of fledglings (rs=0.536, n=11, P=0.088).

Figure 5.

Mean clutch size (±SD) of reed warbler in study years. Number of nests sampled each season ranged from 30–125.


Over the past 37 years, the breeding season of reed warblers in our study population has changed dramatically. Breeding commences 3 weeks earlier than it did in 1970 and the median first egg date has advanced by over 2 weeks. This phenomenon has been reported for other species (Sanz 2002a, 2003, Schaefer et al. 2006), although there are some groups that have not advanced their phenology in response to climate change (Parmesan and Yohe 2003, Visser et al. 2003).

The onset of laying correlated significantly with the mean temperatures in April–May. Similar relationships have been found in some other studies (Both et al. 2005b, Schaefer et al. 2006), and were attributed to temperature-related changes in vegetation and insect development (e.g. Hawkins and Sweet 1989). However, our observations suggest that the advancement in reed phenology was stronger than the advancement in the start of laying, and hence reed warblers are now late considering reed phenology. In the 1970s and the 1980s (with the exception of 1983), most early nests were built in completely dry reeds, as new reeds were too short (Borowiec 1985). In contrast, during the two last decades (and the warm season of 1983), the first nests were built in a mixture of dry and fresh reeds (Borowiec 1985, Halupka and Dyrcz unpubl. data). It might be advantageous for the species to be late in relation to reed development. During seasons with warm springs, early nests were better protected by being hidden in newly emerged reeds. As a result, these nests suffered fewer losses from predation (Borowiec 1985, Halupka unpubl. data, cf. Schulze-Hagen et al. 1996).

It is possible that reed warblers will further advance their start of breeding in the future. However, although first males arrive about 2–3 weeks earlier compared to the1980s, the arrival of first females is much less advanced (Borowiec 1992, Witkowski et al. 1995). Hence, the advancement of laying is probably constrained by female arrival time (cf. Both et al. 2005a). They, in turn, may be constrained by other factors, e.g. the species’ endogenous rhythms, climate in wintering areas, etc. (Both and Visser 2001).

A shift in laying phenology associated with climate change has been reported in several other studies (Both and Visser 2001, Sanz 2003, Schaefer et al. 2006) and generally attributed to the synchronisation of chick production with peak occurrence of food abundance (Blondel et al. 1993, Both and Visser 2001, Cresswell and McCleery 2003). The previously studied species, such as tits and flycatchers, feed their nestlings with caterpillars, which are an ephemeral resource. Therefore, any mismatch between the peak of caterpillar abundance and the peak of food provisioning leads to a decrease in the production of young and/or their physical condition (Visser et al. 1998, Both and Visser 2001, Thomas et al. 2001). In reed warblers, such a situation does not exist. It has been demonstrated that the species feeds the nestlings various kinds of food (Dyrcz 1979, Bibby and Thomas 1985, Schulze-Hagen 1991). In our study area, no clear peak of food abundance was detected (Dyrcz and Zdunek 1996). In contrast, food resources were found to be rich throughout the season and nestling starvation occurred only in bad weather conditions (Dyrcz 1981, Borowiec 1994, Halupka and Wróblewski 1998).

Median first egg dates correlated significantly with mean seasonal temperature: in warmer seasons, the peak of laying took place sooner. The outlying point in Fig. 3 comes from 1980, when, in spite of a very low May–July temperature (14° C), the median first egg date was still relatively early. We attribute the outlier to very bad weather conditions in July, when precipitation was the heaviest compared to other study years. Under such wet conditions, nestling mortality is very high (Dyrcz 1979, 1981, unpubl. data), and birds restrain from further laying. We found a significant negative correlation between the end of laying and July precipitation (rs=−0.737, n=12, P=0.013), but not temperature (rs=−0.149, n=12, P=0.654). In 1980 birds ended laying extremely early, on July 10, which shifted the median first egg date. Heavy July rains in 1973, 1980, and 1981 similarly encouraged an early end to laying, which resulted in short laying periods (outlying points in Fig. 4).

While many studies have found that species have shifted their entire breeding season, both starting and ending the season earlier, the end date of breeding for reed warblers has remained unchanged. As a result, their breeding season is longer overall. To our knowledge, this finding is the first of its kind. The lengthening of the laying period by about 3 weeks has important implications for species reproduction. This constitutes an important part of the nesting episode, which lasts 26–27 d (from laying of the first egg till fledging of the last young). Given a longer laying period, more birds are able to rear second broods. In the 1970s and the 1980s, only a few early-mating pairs had a chance to produce second broods, provided that the first was successful. Overall, only about 0–15% of individuals laid second clutches (Dyrcz 1981, Borowiec 1992). Between 1994 and 2006 up to 35% of birds reared second broods (Halupka and Wróblewski 1998, unpubl. data).

Though clutch size is strongly related to the population's laying date (Dyrcz 1981, Borowiec 1994, Halupka and Wróblewski 1998, unpubl. data), the advancement in laying was not associated with a significant increase in clutch size, as some authors have both predicted (Sanz 2002b, Winkler et al. 2002) and found (Järvinen 1989, Winkel and Hudde 1997; but see Bairlein and Winkel 2000, Winkler et al. 2002). However, nesting success tended to increase with increasing temperature, a result also found in the great reed warbler Acrocephalus arundinaceus (Dyrcz 1981). At least two reasons may be responsible for this relationship: (1) in warmer seasons, early nests are better hidden and suffer fewer losses (Borowiec 1985, Schulze-Hagen et al. 1996), and (2) in warmer seasons, insects are more abundant and food chains more complex, which may decrease predator pressure on nests.

It would appear that the studied population of reed warblers benefits from climate warming. First, the longer breeding season results in increased possibilities for re-nesting. Second, in warmer years, early nests are better protected and thus nest losses are lower. Both factors should affect the total number of nestlings produced by a breeding pair during the entire breeding season and lead to increased fitness.


We would like to thank Ewa Dabrowska, Jan Lontkowski, Maria Podzorska, Jolanta Wójcik, Jaroslaw Wróblewski, Wanda Zdunek and many other for field assistance. We are extremely grateful to two anonymous reviewers whose detailed and thorough remarks helped us to considerably improve the manuscript. The text benefited from the comments made by Konrad Halupka. Jessica Pearce, Margaret Fazio and one of the referees helped us improve the English. In some study years the research was financially supported by the Polish State Committee for Scientific Research (grants no 6P20403936 and 2P04F05330).