26 April 2013

Wintering Sand Martins benefit from fewer conspecifics

The theme of this Blog (Demography - "The study of the characteristics of populations, such as size, growth, density, distribution, and vital statistics") certainly applies to David Norman’s long-running study of Sand Martins breeding in mid-Cheshire. In more than 200 mist-netting sessions over 23 years (1981-2003), he ringed 6,053 adult Sand Martins and recaptured 1,179 of them in a subsequent year, many of them multiple times for a total of over 10,000 handlings of adult birds.

Mist netting - David Norman

This long run of data with large sample sizes, and annual assessments of the breeding population from counting nest-holes, allowed a detailed analysis, jointly with Will Peach, which has recently been published in Ibis. The details are in the paper (click here for a copy or e-mail david@davidnorman.org.uk) but in outline the work shows …

1. The annual survival of adult Sand Martins averages around 35%, varying from as low as 10% to a high over 60%, mostly determined by the rainfall in their Sahel wintering quarters. There are fewer insects in the dry years and more birds starve. This is already well-known for several trans-Saharan migrants but this study also showed that the effect is non-linear: above a certain level of African rainfall the birds’ survival flattens off, limited by mortality elsewhere in the life cycle.

2. In this study, there was no effect of summer weather (temperature or rainfall) on adult survival.

3. The size of the breeding population is mostly determined by the survival and return of adults, and much less by recruitment of new birds (one-year-old first-time breeders and immigrants from elsewhere).

Sand Martin - Lawrence G Baxter

 4.  This is the first study to show that overwinter survival in the Sahelian winter quarters is density-dependent. Thus, if the population is high, there is more competition for insect food and more Sand Martins die; if the population is low, even in a dry year there is more food to go round and more martins survive and return to the breeding grounds. 'Population' here means the winter population of all western European Sand Martins, which mix in the Sahel during winter.

5. The recruitment rate of first-time breeding adults was also density-dependent. More un ringed ‘new’ adults were captured during summers when local colony size in Cheshire was relatively small, and vice versa. Competition amongst breeding pairs for nesting sites or food might have caused this pattern.

Sand Martin - John Harding

David and Will comment that such density dependence was suspected from previous studies on other species but required a vast amount of fieldwork and statistical analysis to prove it. Such competition for insect prey may apply to other insectivorous migrants that rely on the seasonal flooding of wetlands across the Sahel zone. Better rains in recent years have allowed breeding populations of several Sahel-dependent species to increase in recent years, but the threat of drought continues to hang over the people and the wildlife of the region.

Thanks to David Norman for letting us know.


  1. David is to be congratulated on his perseverance in gathering the data for this paper, and with Will, for getting it published in Ibis.

    For me, the main interest is not the sorting out of the density-dependent issues, much work as this must have taken, but his findings on the composition of the breeding population. Clearly, I am not understanding what I am reading in this paper, although this demography team article, paragraph 3, suggests that others have read it as have I.

    The paper says:

    “Inter annual variation in adult survival therefore had a much greater influence on population size in Cheshire Sand Martins than variations in fecundity”
    “suggests that population size in Cheshire Sand Martins was determined by inter-annual variation in adult survival with recruitment having a weaker influence”

    and to me this means that the breeding population was mainly composed of surviving adults as they “had a much greater influence on population size”, with recruitment supplying the smaller proportion, “with recruitment having a weaker influence” but clearly this was not the situation that David found in the field. With an overwinter survival rate averaging 35% during the early years of his study (prior to 1995) and recruitment having a weaker influence, his population would have fallen annually during this period, or at least, fallen overall during this period. (He found that the size of the local breeding population fell in all but 5 years when survival rate was below 35%). Instead of which, the population rose c1300% from 90 in 1984 to 1236 in 1996. Of course, the Sand Martin population in Britain is noted for its sometimes wide swings from year to year but in David’s Study, even in the years following the best survival rates, with recruitment having a weaker influence, there could be no great annual increases in the population level.

    I await his presentation at the Annual Conference to find where I am misunderstanding the situation.

    1. We thank Ted for his comments on our Sand Martin study. We accept his point that annual recruitment must be important in any bird population with an adult survival rate of 0.35. Clearly, those 65% of adults that die each year need to be replaced by new recruits if population size is to be maintained. Thus in an absolute sense, recruitment is more important. The point we made in the paper was that year-to-year changes in the size of our population were more strongly correlated with variation in adult survival than with variation in the estimated recruitment rate (note that our definition of recruitment is the number of new recruits in the population in any year per adult female present in the previous year). This suggests that temporal variation in adult survival had more of an influence on population size than did temporal variation in recruitment (see Figure 3 in the paper). In this context of understanding population change, it is not necessary for the most important demographic driver of population change (adult survival in this case) to account for the majority of breeding adults in any year; it is the extent and pattern of variation over time that matters.

      We accept that our demographic estimates do not provide a good explanation of the large increase in Sand Martin breeding numbers that we recorded between 1994 and 1996 (from 445 to 1236, or 178%). Although adult survival was relatively high during 1994-95 (see Figure 1 in the paper), our estimates of recruitment and adult survival would (when combined) predict lower population growth (62%) than we actually observed over that two-year period. This suggests that one or both demographic rates were underestimated during that two-year period.

  2. I spoke to David at the BTO Conference and he intends to answer my queries here. I hope, however, that this may remove the necessity for that. I knew that I was missing something that I should not be, and because of the quality of the Authors of this paper I have spent a deal of time looking for what it could be. For me, analyses are a necessary evil and whatever they throw up must relate to reality.
    Keeping this as short as practicable (Refs. are available), The Sand Martin is considered to be douuble brooded as a rule but there are annual variations in the degree of double broodedness. In order to examine another approach I looked at the present day situation.
    In the 18 years from 1996 to 2013 inclusive, at a site in southern Scotland which was/is kept under daily observation helped by site and nest cameras running continuously, 2164 successful broods were produced from 1279 holes. 417 holes (0.326%) held one brood, 839 (0.656%) holes held two broods and 23 holes (0.018%) held three broods.
    If the most experienced birds (a male normally owns a hole for a season, or longer) could be expected to produce the most broods, then to apply David's 42% mean survival rate, (higher than most have found) to support the findings as much as possible, then 537 experienced birds at the site above would have produced 1097 broods ( 23x3 and 514x2) and inexperienced birds 1067 broods (417x1 and 325x2).
    It would seem, then, that the experienced birds do have it, except that the survival rate of brood members falls for broods fledging later as the breeding season progresses and many second brood members are markedly less successful than first broods ones, with the few third brood members even more so.
    Giving all the support possible to the particular finding under discussion (too much, really) this method finally fails to support it, but it reinforces my belief that though experienced breeding birds are normally a minority in the population, in any given year they are the mainstay, with the majority of the breeding population, the first year breeders, a froth that will provide the mainstay, but minority of the following year.
    Of course, I might be wandering along an unexplored pathway so far as the analyses went, but at least it has made me look once again at this particular matter.

    1. Ted’s interesting modelling does indeed demonstrate a possible way through which annual survival of adults could dominate the demography of Sand Martins, if experienced adults are substantially better at producing fledglings that survive to the next breeding season. Our work, which concentrated on retrapping adults in the first half of the season, cannot shed any light on this.

      Our analysis suggested that an annual (male) survival rate of 0.35 is a possible threshold level of adult survival for population growth: the size of the local (Cheshire) breeding population fell in all five years when male survival was below 0.35 and increased in eight of the 11 years when male survival exceeded 0.35.

      Survival varies enormously between years (with extremes of 0.11 and 0.66 in our study), making it difficult to derive an average figure for survival rate. We found that adult male survival averaged 0.32 during 1981-91 and 0.42 during 1992-2001. Ted Cowley’s earlier study found that adult survival averaged 0.19 during 1968-73 and 0.42 during 1973-78: these figures were for unsexed adults, implying, as male apparent survival is higher than for females, that male survival would have been higher than 0.42 during 1973-78.

  3. I thank David and Will for their replies.
    I hope that these comments have been of value to others apart from myself.