[Update: thoughts in this post contributed to Kristensen et al. (2015), which found mathematical conditions under which phenology may delay.]
Climate change has caused an advance in phenological events in many species. In migratory birds, the effects of warming flow causally up the trophic levels. For example, warmer temperatures lead to earlier plant phenology (e.g. budding), which leads to earlier peaks in the abundance of foods (e.g. insect larva) that are important to raising nestlings, which puts pressure upon birds to advance their own breeding timetable. In general, birds have responded to warming weather by advancing their own phenology. Migratory birds are both arriving earlier and laying earlier in recent years. However, the responses have not been uniform.
Generally we expect that migratory birds will arrive earlier with warming temperatures. However there have been a few studies in which a delay in arrival time has been reported as a response to advancing phenology (Mason 1995, Huin & Sparks, Sparks & Braslavsk ́2001, Penuelas et al. 2002, Butler 2003, Cotton 2003, Gordo et al. 2005, Primack et al. 2009, Swanson & Palmer 2009, Ellwood et al. 2010, Lee et al. 2011, Gordo & Doi 2012). These studies typically look to factors like e.g. temperatures in other parts of the year as an explanation (Gordo et al. 2005, Swanson & Palmer 2009).
Ogawa-Onashi & Berry (2013), in a paper emphasising the need to integrate international perspectives in this area, summarise the situation as follows:
In contrast to plant phenology that shows a general advancing trend in spring, changes in animal phenology are variable between species and locations (Kamitani, 2010a and Primack et al., 2009a). However, by examining six plant and six animal species over the period 1953–2005 at up to 160 observation sites across Japan and Korea, Primack et al. (2009a) showed spring phenology of all plants advanced while that of all animals was delayed at over half the sites. Moreover, the first singing date of Cettia diphone riukiuensis (Ryukyu bush warbler) was delayed by 2.6 days/decade between 1953 and 2005, which was caused by male population declines due to human population increases (Gordo and Doi, 2012). Similarly in South Korea, monitoring of Hirundo rustica (barn swallows) between 1971 and 2008 showed that they are arriving on average 10 days later, with population declines also suggested as a possible reason (Lee et al., 2011). On the other hand, delays in the first appearance of Orthetrum albistylum speciosum (common skimmer) (4.8 days/decade) was not influenced by human population density, but by shifts in dragonfly life cycles (Doi, 2008). Further, delays in the first appearances of Pieris rapae (small white) were caused by winter warming that delayed timing to meet the chilling requirements for breaking dormancy (Kamitani, 2010a). Thus, although the causes of phenology delays vary among species, these studies clearly indicate that delayed spring events are widespread in Japan for birds and insects, which are contrary to the trends observed elsewhere (e.g. Bertin, 2008 and Parmesan, 2007).
A key explanation for these observations mentioned above is that they are the result of population declines; specifically, the result of sampling error induced by the declines. As the population declines, even if the mean arrival time does not change, the likelihood of observing the first arrival will decline (Miller-Rushing et al. 2008) (also Huin & Sparks). All of the above studies used first observed arrival date (FAD) as the measure of arrival time, therefore it is possible that their anomalous result is a consequence of this sampling error, and recently delay in FADs has been successfully used as a proxy for detecting population declines (Lee et al. 2011, Gordo & Doi 2012).
Ideally one would find a delay reported from whole-profile arrival times. I have looked and have not been able to find such a study. The most interesting I can find is the study by Lee et al. (2011), who observed delayed arrival time in barn swallows in Korea. The study simulated the population decline that would have been necessary to cause such an observation artifactually, and estimated extreme declines of 99%. However, comparing this to similar populations where the decline was known, they believed that the estimated decline - though extreme - is plausible.
The dramatic delays in arrival dates seen here in Korea are most likely explained by declines in population size of 99% or more. While this seems to be an extreme decline, the Chungbuk site had 93% decline in swallow populations in just 10 years and the majority of Korean observers report a 50% or more decline of swallows at their sites. So these estimates of population decline may be accurate. However, this still needs to be confirmed by the discovery of past field surveys at other sites combined with a current re-census to document the decline in population size. Alternative explanations, such as changes in migration pathways and changes in the number of broods raised per season should also be investigated.
Short of those alternative explanations panning out, it would appear that the observation of delayed arrival is an artifact of sampling on declining populations. However the fact that this kind of result is relatively common in Asia (Ogawa-Onishi & Berry, 2013) is intriguing. Returning to their main study, Primack et al. (2009) examined 6 plants and 6 animals, and found that 5 (all plants) had advanced phenology, 5 (all animals) had delayed phenology, and 2 had mixed responses. They were at that time sure that the delays were unlikely to be the result of population declines
delays in the timing of spring activity can be caused by declining population sizes, which can shorten the duration of the event but not affect the mean value (Tryjanowski and Sparks, 2001, Miller-Rushing et al., 2008a and Miller-Rushing et al., 2008b); or could just be due to the fact that densities are lower and the probability of observing the event when it really first happens is lower. Such declines are occurring for many animal species in Japan (Higuchi, 1996 and Lane and Fujioka, 1998). Nevertheless, we found that there are dramatic differences among species in their phenological responses to changes in temperature (Table 3, Fig. 2), responses that should be minimally affected by changes in population size.
They correctly predicted the follow-up studies in Ogawa-Onashi & Berry (2013), that the results are idiosyncratic, in their observation that generalisations made for a species at a particular location don’t translate to other locations:
Due to the site-specific conditions created by not only temperature but also by many other variables characteristic of the site, species might be able to track the changing climate in some locations but not in others. The assessment of such differences will be critical for conservation efforts.