1.4.1 Larger-scale aggregation
This section evaluates studies that use techniques that aggregate from individual observations at sites to regional, continental and global scales. Meta-analysis is a statistical method of combining quantitative findings from many studies investigating similar factors for the purpose of finding a general result. The methods used in the various studies, however, need not be similar. The criteria for inclusion of studies in a meta-analysis are determined a priori, and rigorously followed to avoid investigator effect.
Several studies have examined the ‘fingerprint’ of observed warming in recent decades on the phenology and distribution of plants and animal species using meta-analyses (Root and Schneider, 2002; Parmesan and Yohe, 2003; Root et al., 2003). Although the detailed results of these studies are different, because they used different species and different methods, they all conclude that a significant impact of warming is already discernible in animal and plant populations at regional and continental scales in the Northern Hemisphere.
One meta-analysis (Parmesan and Yohe, 2003) of 31 studies of more than 1,700 species showed that recent biological trends matched the expected responses to warming. They estimated northward range shifts of 6.1 km/decade for northern range boundaries of species living in the Northern Hemisphere and advancement of spring events in Northern Hemisphere species by 2.3 days/decade. They also defined a diagnostic fingerprint of temporal and spatial ‘sign-switching’ responses uniquely predicted by 20th-century observed climate trends. Among long-term, large-scale, multi-species data sets, this diagnostic fingerprint was found for 279 species. They concluded, with ‘very high confidence’, that climate change is already affecting living systems.
After examining over 2,500 articles on climate change and a wide array of species from around the globe, another study found that 143 studies fitted the criteria for inclusion in their meta-analyses (Root et al., 2003). They focused on only those species showing a significant change and found that about 80% of the species showing change were changing in the direction expected with warming. The types of changes included species expanding their ranges polewards and higher in elevation, and advances in the timing of spring events by about 5 days/decade over the last 30 years. This number is larger than the 2.3 days/decade found by Parmesan and Yohe (2003), because those authors included both changing and not-changing species in their analysis, while Root and co-authors only included changing species. A more recent meta-analysis of bird arrival dates (Lehikoinen et al., 2004) showed strong evidence of earlier arrival. Of 983 data series, 39% were significantly earlier and only 2% significantly later for first arrival dates.
The EU COST725 network analysis project had as its main objective the establishment of a comprehensive European reference data set of phenological observations that could be used for climatological purposes, particularly climate monitoring and the detection of changes (see Box 1.3).
Box 1.3. Phenological responses to climate in Europe: the COST725 project
The COST725 meta-analysis project used a very large phenological network of more than 125,000 observational series of various phases in 542 plant and 19 animal species in 21 European countries, for the period 1971 to 2000. The time-series were systematically (re-)analysed for trends in order to track and quantify phenological responses to changing climate. The advantage of this study is its inclusion of multiple verified nationally reported trends at single sites and/or for selected species, which individually may be biased towards predominant reporting of climate-change-induced impacts. Overall, the phenology of the species (254 national series) was responsive to temperature of the preceding month, with spring/summer phases advancing on average by 2.5 days/°C and leaf colouring/fall being delayed by 1.0 day/°C.
The aggregation of more than 100,000 trends revealed a clear signal across Europe of changing spring phenology with 78% of leaf unfolding and flowering records advancing (31% significantly (sig.)) and only 22% delayed (3% sig.) (Figure 1.6). Fruit ripening was mostly advanced (75% advancing, 25% sig.; 25% delayed, 3% sig.). The signal in farmers’ activities was generally smaller (57% advancing, 13% sig.; 43% delayed, 6% sig.). Autumn trends (leaf colouring/fall) were not as strong. Spring and summer exhibited a clear advance by 2.5 days/decade in Europe, mean autumn trends were close to zero, but suggested more of a delay when the average trend per country was examined (1.3 days/decade).
The patterns of observed changes in spring (leafing, flowering and animal phases) were spatially consistent and matched measured national warming across 19 European countries (correlation = -0.69, P < 0.001); thus the phenological evidence quantitatively mirrors regional climate warming. The COST725 results assessed the possible lack of evidence at a continental scale as 20%, since about 80% of spring/summer phases were found to be advancing. The findings strongly support previous studies in Europe, confirming them as free from bias towards reporting global climate change impacts (Menzel et al., 2006b).
Figure 1.6. Frequency distributions of trends in phenology (in days/year) over 1971 to 2000 for 542 plant species in 21 European countries. From Menzel et al. (2006b).