Climate Change? A Statistical Account of Flowering in East Lothian: 1978-2001

Fred Last*, Adrian Roberts & Desmond Patterson

From the 1960s onwards, there has been a growing unease in much of the world about the increasing impact of human activities on the environment. This was initially focussed (in the 1960s and 1970s) on the increase in ‘acid rain’ and its effects on freshwater ecosystems and the growth of trees, but towards the end of the 20th century, the emphasis changed. With the identification of the ozone hole over Halley Bay, Antarctica in 1985, attention turned to the effects of greenhouse gases (‘man-made’ carbon dioxide, oxides of nitrogen, ozone, marsh gas (methane) and CFCs). These gases have the ability to retain solar energy re-radiated from the surface of the earth, so enhancing the greenhouse gas effect attributable to the smaller concentrations of natural carbon dioxide, methane etc which maintain the surface of the earth at 15°C instead of -6°C, which it would otherwise be.

There are many potential influences of climate change. Near to home, two of them are of undoubted significance:

a) the increasing levels of the oceans, concomitant with the melting of the Antarctic and Arctic ice caps, may result in the submergence of parts of the Forth estuary

and – more importantly for this essay

b) the anticipated increase in air temperatures that would be expected to affect plant performance – but how would the changes in plant performance be manifested?

For three centuries, the study of events (phenology), particularly first events, has had its adherents. In 1736, the Marsham family in Norfolk initiated, in a wide-ranging study (Sparks, TH & Carey, PD 1995 pp321-9) a series of observations of the first dates of flowering of four species of plants – hawthorn, snowdrop, turnip and wood anemone. These observations were maintained by successive generations of the Marsham family until 1947. During this period the first dates of flowering of hawthorn and snowdrop remained unchanged: that of turnip was delayed while wood anemone flowered progressively earlier (0.10 days per year). Soon after the start of the Marsham records, Gilbert White (White, G c1780s) of Selborne, with the encouragement of Daines Barrington, embarked upon a series of daily records including many related to flowering. These were given in White’s annual series of Naturalist’s Journal that terminated at his death in 1793. The interest arising from this elegant series of very simple observations, needing nothing more than pencil and paper, and those of the Royal Meteorological Society (Jeffree, EP 1960 pp95-103), of Willis in Norwich (1913-42) (Smith, LP 1968) and more recently the Fitter (Fitter, AH & Fitter RSR 1995 pp55-60; 2002 pp1689-91) family at Chinnor, Oxfordshire (1954-89) encouraged the interest of the senior author with his longstanding interest in the effects of atmospheric pollutants.

From early 1978, he, his late wife Pauline, and his father monitored flowering in their East Lothian garden (between Prestonpans and Longniddry). With relatively few gaps, observations of 600-1000 plant species (mostly biennials and perennials) were made at weekly intervals to the present – a total of not less than 1.2 million pieces of information. To minimise the risk of bias it was decided to collect data for at least 20 years before embarking upon the process of analysis. Flowers were judged to be open when stamens and/or stigmas could be seen without parting petals or sepals.

In 2001, Scottish Natural Heritage (SNH) commissioned Biomathematics & Statistics Scotland (BioSS) to make preliminary analyses of the data that had been accumulated (Roberts, AMI; Last, F & Kempton, E. in press). In common with most people we assumed that the onset of flowering, in the second half of the year, of late flowering species might be affected significantly by summer weather – in the event this didn’t prove to be the case. By averaging the first dates of flowering of plants that usually started to flower in January/February, March, April, May, June, July or August/November it was found that year-to-year variations were of a similar nature irrespective of flowering time (Figure 1). The mild January/March of 1989 proved to be ‘early’ for plants that usually started to flower in March, also for those flowering in August /November, while 1996 proved to be a late season for all species. In short there was a very strong imprint of season. However, there was much more variation in time of flowering in the first quarter of the year than in the second half.

The variations in the first dates of flowering of early flowering species were mostly very much larger than we had expected (Figure 2 and Figure 3). In mild, early seasons, the first dates of flowering were commonly ten or more weeks earlier than in ‘late’ years. For example, snowdrops started to flower on 9th January in 1979 (an early year) but on 12th March in 1984 (a late year) a range of nine weeks (62 days). For the wild daffodil (Narcissus pseudonarcissus), the range was 45 days (6.4 weeks), from 14th February in 1979 to 31st March in 1984, but with Mexican Orange (Choisya ternata) the range, from 6th February in 1989 to 25th May in 1984, widened to 108 days. The largest range occurred in the poached egg plant (Limnanthes douglasii), an astonishing 180 days, virtually six months. These observations indicate that the magnitude of seasonal responses differed markedly between species. As a result, the sequences of plants coming into flower varied from year to year (Figure 4). In an early season, Mexican Orange started to flower before Skimmia, which in turn started before Honesty (Lunaria annua); but in a late season Honesty, which is one of the least variable species of plants, started before Skimmia which preceded Mexican Orange – a complete reversal, reminiscent of the rhyme

Ash before oak, We’re in for a soak.
Oak before ash, We’re in for a splash.

With so much variation observed from year-to-year it was not expected that long-term changes suggestive of an effect of climate change on first dates of flowering would be found. However, when the trends for each of 173 species, with data spanning at least 20 years, were plotted against their mean first dates of flowering, a significant linear relationship was identified (Figure 5). It showed a marked trend towards earliness, approaching one day per year, among species that started to flower between January and March inclusive. This trend, based on a mix of British native and introduced species, applied equally to native species when analysed separately (Figure 6). The present trend, about one day per year, is ten times faster than that (0.1 day per year) calculated for wood anemone using the Marsham records.

Some of the trends for individual species are given in Figure 3. Of the nine species depicted, all except those in the Rosaceae family (apple and Rosa ‘Nevada’) have trends to increasing earliness. Interestingly, Figure 3 suggests that the members of the Rosaceae family behave distinctively from those of other plant families. While Mexican Orange and most other species flowered early in 1989, the responses of rosaceous plants were inexplicably delayed until 1990.

It is one thing to describe what has happened but how can the different seasonal patterns and long-term trends in flowering be explained? To try to answer this we, like those before us, have focussed on the possible influences of environmental variables such as air and soil temperatures, and also rainfall.

These environmental attributes have been stressed in a recent briefing Report (Climate Change Scenarios for the United Kingdom: The UKCIP02 Scientific Report, 2002), commissioned by The Department for Environment, Food and Rural Affairs (DEFRA) as part of the UK Climate Impacts Programme. This Report points out that the mean annual air temperature of central England temperature has risen by almost 1°C, and that the 1990s was the warmest decade since records began in the 1660s. The Intergovernmental Panel on Climate Change (IPCC Third Assessment Report – Climate Change 2001)) concluded that

most of the warming observed over the last 50 years is likely to have been due to increasing concentrations of greenhouse gases.

Interestingly these Reports strongly suggest that in many areas night-time temperatures have increased more than day-time temperatures, suggesting that our climate is becoming less cold at night rather than warmer during the day.

With these wide-ranging generalisations to guide us, what has been happening in the Lothians? Can the changing patterns of first flowering be related to temperature changes and/or those of rainfall? The daily records taken at the Royal Botanic Garden, Edinburgh (1977 – 2001) show that the warmest and coldest years were 1997 (13.2°C mean yearly maximum) and 1979 (11.6°C) respectively; the wettest and driest were 1990 (yearly mean of 2.1mm per day) and 1996 (1.7mm) and the sunniest and dullest were 1995 (yearly mean of 4.3hr per day) and 1985 (3.5hr per day). Air temperatures were maximal in July and minimal in January. The monthly minimum and maximum air temperatures were 1°C and 6.5°C in January and 11.3°C and 18.9°C in July. Significantly, air temperatures, but not rainfall and soil temperatures, increased appreciably between 1976 and 2001 (Figure 7). The mean maximum air temperature increased by nearly 0.67°C, or 1.0°C, in 37 years, the increase being more evident during the winter months than at other times of the year.

When bringing together the temperature records and those of first dates of flowering the majority of plant species showed statistically significant relationships indicating that

the first dates of flowering advanced in the warmer years; and
irrespective of the dates of flowering, the advance was largely attributed to temperatures in January, February and March of the current season, also to May temperatures for those plants starting to flower in May or later.

Thus the onset of flowering in March was strongly associated with air and topsoil temperatures of the current month (March) and those one (February) and two (January) months earlier. The onset of flowering in plants that usually flower between August and November was strongly associated with the mean monthly temperatures up to seven months previous (January, February and March); there were no similar systematic associations with attributes such as rainfall and sunshine. These conclusions support the evidence presented in Figure 2, which suggests that the flowering imprint of the year was to a very large extent determined by January to March temperatures.

What about the long-term trend showing that many early (January, February and March) species are now (2001) flowering nearly three weeks earlier than was the case in 1978? The analyses suggest that the trend is significantly associated with long-term temperature increases which, as already mentioned, were found to be greater in January, February, and March than at other times of the year. As these temperature increases have been shown, with conviction, to be related to the abundance of greenhouse gases in the atmosphere it would appear that the 24-year run of data collected in East Lothian (even though relatively short in terms of natural cyclical changes), are sufficient to lend support to the contention that East Lothian’s climate, as elsewhere in Scotland (Scottish Executive Central Research Unit 1999), has changed enough to affect at least some aspects of plant performance.

Is this a matter of concern? Should there be more concern, recognising that the effects of increasing temperatures on first flowering may be supported in due course by changes to other aspects of plant performance such as the duration of flowering and the numbers of flushes of flowers per year – attributes that relate to the vegetative and reproductive phases of plant growth – and the seasonal availability of nectar, foliage and seeds. But, perhaps even more importantly, what will happen if the conditions favouring earliness continue? The present rapid trend, noted for some species, of one day per year cannot continue – it is a biological impossibility.

At international conferences during the 1990s, politicians have grappled with the climate change issue with information less than convincing to the general public. However, their task should be eased with the steady accumulation of objective data and the perceptible softening of public attitudes to the acceptance of the precautionary principle. The changes recorded in Longniddry, if allowed to develop unchecked and if projected to the performance of natural ecosystems, agriculture, forestry and more widely to horticulture, lead to the conclusion that steps taken to minimise the emission of greenhouse gases, most notably an integrated energy and transport policy, are likely to be richly rewarded.

Acknowledgements

Thanks are due to Ms Emily Kempton (BioSS); Dr Noranne Ellis (SNH); and Nick Cheales and Gill Calder, both of the Scottish Wildlife Trust, for their willing help with this essay.

Fred Last*, Adrian Roberts & Desmond Patterson

Acknowledgements

Further reading & references