Due to the scarcity of instrumental climate records prior to the XXth century, estimates of global climate variability during past centuries must rely upon indirect “proxy” indicators. A proxy indicator is a local record (natural or documentary) that is interpreted using physical or biophysical principles to represent some combination of climate- or environmental-related variations back in time. Palaeoenvironmental proxies have the potential to provide evidence for large-scale climatic changes prior to the existence of instrumental or historical documentary records. Careful calibration and cross-validation procedures are necessary to establish a reliable relationship between a proxy indicator and the climatic variable or variables it is assumed to represent, providing a “transfer” function through which past climatic conditions can be estimated. High-resolution proxy climate indicators, including tree rings, corals, ice cores, and laminated lake/ocean sediments, can be used to provide detailed information on climate variations back in time, which can be complemented by certain coarser resolution proxy information (e.g. boreholes and non-laminated ocean sediment records). Some of the most commonly used proxy indicators are outlined below, paying special attention on tree-rings, due to their relative importance in the context of this Thesis.
Historical climatology is based on the study of different kinds of references that can be associated to climate events. For some areas, written evidences are available for several millennia and, although an explicit record of climate variables is not usual, they are often referred indirectly. Thus, references to the effects of flood events, long periods of drought or extremely cold winters can be found in religious archives in the form of rogation ceremonies, as well as in agricultural records (Barriendos and Martín-Vide 1998; Bell 1975; Pfister et al. 1999).
Ice cores from polar regions and mountain glaciers provide several indicators of past climates, including stable isotopes (d18O), the fraction of melting ice, accumulation rates, concentration of different salts and acids, or the amounts of pollen or trace gases (e.g. CH4 or CO2) in the atmosphere, among others (Alley et al. 1997; Francey et al. 1999). Ice core data has a high temporal resolution (from annual to seasonal), and provides simultaneous information from climate-related and atmospheric variables, that might be helpful to assess the effect of CO2 increase in global climate. The best dated series are based on sub-annual sampling of cores and the counting of seasonal ice layers, diving absolute errors of just a few years in a millennium (Folland et al. 2001). Ice cores have the disadvantage of being located far from the areas of human devolopment. This is a strength from the point of view of global climate issues, as they are less affected by local anthropogenic effects. However, it is also a disadvantage in order to assess the role of past climate changes in human evolution.
The particular biochemical composition of pollen grains makes them relatively resistant to chemical, biological and physical damage. Consequently, pollen grains may survive millions of years in a large variety of sedimentary environments: peat bogs, lake and marine beds, several kinds of loose terrestrial sediments, and even in consolidated rocks (for example in stalagmites). As pollen deposition, for a given species, is expected to be proportional to its abundance, palynologists are able to reconstruct past vegetation from the study of fossil pollen assemblages. Furthermore, and provided a good knowledge of the ecological range of a given taxa, functional group or plant community, it has been possible to derive climatic information from pollen data (Lebreton et al. 2004; López-Sáez et al. 2003). Nevertheless, it should be noted that anthropogenic influence on vegetation has increased exponentially since the onset of agriculture. Thus, during great part of the Holocene, pollen data might have too human-derived noise to provide reliable climatic information, although still giving useful information about the landscape (Committee on abrupt climate change 2002; Lebreton et al. 2004). Indeed, most pollen assemblages for the last two millennia reveal clear symptoms of anthropogenic disturbance, such as the abundance of pollen from cereals and other crops, and the spread of ruderal species (Davis 1994; Lebreton et al. 2004; López-Sáez et al. 2003).
Annually laminated (varved) lake sediments also provide high-resolution records of palaeo-environment. When annual deposition of varves can be independently confirmed through other dating techniques (e.g. radiocarbon) they provide seasonal to inter-annual data. Three main climate variables may influence lake varves: summer temperature, winter snowfall and rainfall. Moreover to sedimentological studies (grain size, sedimentation patterns), lake sediments can be analysed from a variety of approaches: isotope analyses (d13C in organic matter, as a reflect of the dominant biota; d18O in carbonates and shells as indicator of evaporation rates), chemical analyses (changes in salt composition, abundance of different organic compounds, trace elements), pollen records, diatoms analyses (Davis 1994; Riera et al. 2004; Rodó et al. 2002).
Ocean sediments also offer high-resolution archives of climate, applying similar methods as those used in lake sediments. However, annually laminated sediments are not usual (Folland et al. 2001). Otherwise, sedimentation rates may still be enough to provide information on a century or millennial scale (Duplessy 2004; Moreno et al. 2004), based on radiocarbon or other external dating, such as volcanic ash shards. The range of variables that can be determined in marine beds is similar to that of continental lakes: sedimentology, stable isotopes in shells and organic matter, chemical analysis, pollen records (Committee on abrupt climate change 2002; Folland et al. 2001).
Annual growths of coral skeletons provide palaeoenvironmental information for tropical and sub-tropical oceans and atmosphere. For example, they have the potential to sample variations in regions sensitive to El Niño Southern Oscillation (ENSO) which can be useful to resolve large-scale patterns of climate (Folland et al. 2001). Accurate age estimates are possible for most sites using a combination of annual variations in skeletal density and geochemical parameters. Palaeoenvironmental reconstructions from corals rely mostly on geochemical variables, such as trace elements or stable isotopes (d13C and d18O) (Juillet-Leclerc and Yiou 2002; Rimbu et al. 2003).
This is a relatively recent methodology, which attempts to provide a direct estimate of ground surface temperatures under certain simplifying assumptions about the geothermal properties of the earth. Although long-term temperature reconstructions have been made for the last two millennia (Bodri and Dovenyi 2004), they are based on assumptions not fully confirmed in such a large-scale, and the temporal resolution of borehole estimates decreases sharply back in time. Thus, borehole estimations of temperature are probably most useful for climate reconstructions over the last five centuries (Folland et al. 2001). Moreover, non-temperature related factors such as land-use changes make often difficult the interpretation of borehole data.
Since the early works of Douglass in 1914 (see (Robinson et al. 1990) for a historical review), tree-rings have been extensively used as palaeoenvironmental proxies. Currently, dendroclimatology is a well-developed subdiscipline of dendrochronology. It is based on the fact that tree growth is often limited by a variety of environmental factors, which may vary depending on the area studied (Cook and Kairiukstis 1990; Schweingruber 1988). For example, in arid and semiarid areas precipitation is the main factor limiting tree growth (Adams and Kolb 2004; Lev-Yadun et al. 1981; Serre 1976), whereas in colder regions temperature becomes the most limiting factor (Briffa et al. 2004; Panyushkina et al. 2003). As the objective is to maximise the effect of a given climate variable on tree growth, that is, to obtain a sensitive tree-ring series, the most useful trees are those found near the margins of their natural growth range (ecotones) where a clear limiting factor can be identificable. Indeed, over an altitudinal transect, for example, we may found significant differences in growth response (Adams and Kolb 2004; Gutierrez 1991; Tardif et al. 2003). As a rule, relatively isolated trees (i.e. without competition) but potentially limited by the desired factor, are the material of choice. For example, if we wish to obtain good inferences on precipitation, trees placed on slopes are preferable than those from the bottom of a valley.
From individual tree-ring width series to large-scale chronologies
Tree growth is not only determined by climate, but is also affected by other, non-climatic factors. Tipically, ring width decreases exponentially with tree age, according to the geometry of an increasing trunk diameter (Kiviste et al. 2002; Rodriguez et al. 2003). However, competition and disturbance effects, among others, might alter this general trend. For example, growth rate of an individual tree can be greatly enhanced after a neighbour tree is cut. Thus, for environmental studies, this “undesired” variability should be statistically removed through standardisation. This involves fitting a curve to the tree-ring series, and then dividing each ring width value by the corresponding curve value. In an ideal tree, fitting an exponential curve would be enough, but in most cases a second detrending is performed through a cubic smoothing spline (Cook and Peters 1981). The spline acts as a high-pass filter, and is defined by a cut-off wavelength. For example, if this wavelength is 20 years, it will remove 50% of variance at this wavelength, with increasing percentages at longer wavelengths (97.5% at 50 years), and lesser removal at shorter wavelengths (2.5% at 8 years). In some cases, further detrending is performed through autoregressive models (Monserud 1986). The final result is always a series of indices with a long-term mean of unity, where higher or lower values for a given year represent proportionally higher or lower growths. To maximise the climate signal, the series obtained from several trees per site are averaged (preferably using a bi-weight robust mean estimation, (Cook and Holmes 1986) to build up a site chronology (Fig. V). The main handicap of statistical treatments is the risk of removing low-frequency climate variability (Cook et al. 1995). In any case, the parameters of standardisation should be carefully determined, always bearing in mind the aim of our particular study.
Stable isotopes in tree-rings
By the time it was first suggested the presence of a climatic signal in plant stable isotopes, tree-ring width analyses were being widely used in palaeoecological studies (Ehleringer and Rundel 1988; Robinson et al. 1990). Hence, the technique was rapidly implemented in tree-rings and several early studies on the relationship between stable isotopes and climate were performed on this material (Craig 1954; Epstein 1979; Gray and Thompson 1976; Libby et al. 1976). Since then, wood or cellulose isotopic composition of different elements has been extensively used in palaeoenvironmental and ecological studies (for further review see (McCarroll and Loader 2004; Switsur and Waterhouse 1998; Warren et al. 2001).
Initially, variations in the d13C of tree-rings was related to temperature (Stuiver and Braziunas 1987; Wilson and Grinsted 1977) and later studies have also reported positive relationships with this variable (Anderson et al. 1998; Saurer et al. 1995). However, according to current models of plant carbon discrimination (see section II.2), little or no direct effect of temperature on d13C is expected. In fact, most of these studies found stronger relationships with other variables, such as relative humidity (Stuiver and Braziunas 1987) or precipitation (Anderson et al. 1998; Saurer et al. 1995; Ferrio et al. 2003a; Ferrio and Voltas 2005). Hence, these results are probably derived from indirect relationships between temperature and plant water status. d13C in tree-rings has shown to be strongly correlated to modelled soil water balance, precipitation, vapour pressure deficit and evaporative demand (Dupouey et al. 1993; Korol et al. 1999; Saurer et al. 1997c; Warren et al. 2001; Ferrio et al. 2003a; Ferrio and Voltas 2005). However, such relationships appear to be restricted to seasonally dry climates: in other contexts, irradiance, altitude and nutrient availability appear to be the main responsive of isotopic variations (Hultine and Marshall 2000; Leavitt and Long 1991; Livingston et al. 1999; Warren et al. 2001).
Early studies on d18O and d2H in tree-rings found that they were strongly correlated with average temperatures, which was initially attributed to the isotopic fractionation in precipitation (Epstein 1979; Gray and Thompson 1976; Libby et al. 1976). However, more detailed physiological studies revealed that the variability associated to evaporative enrichment could not be neglected (Dongmann et al. 1974; Farquhar and Lloyd 1993; Ferrio and Voltas 2005). Thus, the observed correlations with temperature should be attributed to the synergistic effect of this variable on the isotopic signal in precipitation, and on evaporative enrichment (McCarroll and Loader 2004; Ferrio et al. 2005b). On the other hand, evaporative enrichment also explains the strong effect of relative humidity and vapour pressure deficit on d18O and d2H reported by several authors (Burk and Stuiver 1981; Saurer et al. 1997a; Ferrio and Voltas 2005). Nowadays, d18O and d2H signatures in tree rings are known to be determined by three main variables (Roden et al. 2000; Sternberg et al. 2003): the isotopic signature of source water, leaf evaporative enrichment and exchange with xylem water during cellulose and/or lignin synthesis. The latter is determinant for the observed relationship between d18O of tree-ring cellulose and meteoric water, as it enhances the source-water signal, softening the effect of leaf-level enrichment (Barbour et al. 2001; Saurer et al. 1997b; Sternberg et al. 2003). However, in species with a strong stomatal regulation, the original source signal can be lost due to the variability associated to evaporative enrichment (Ferrio and Voltas 2005; Ferrio et al. 2005b).
Although most of the works cited so far were limited to the last 100 or 200 years, isotope studies in tree-ring have been also expanded over some large-scale tree-ring chronologies, providing high resolution climatic reconstructions throughout the Holocene (Becker et al. 1991; Feng and Epstein 1994; Kromer et al. 2004; Libby et al. 1976; McCornac et al. 1994) see further refences in (McCarroll and Loader 2004).
Fig. VI Climate changes in central Greenland over the last 17,000 years, at 100 years intervals. Reconstructions of temperature (black line) and snow accumulation rate (white line). Main climate events are highlighted: 1) Little Ice Age; 2) Medieval Warm Period; 3) Iron Age Cold Epoch; 4) 8k2 episode; 5) Younger Dryas. Adapted from (Committee on abrupt climate change 2002).
After the cold-dry period known as Younger Dryas (ca. 12,800-11,500 cal. BP), begins the longest warm and stable period in the last 400,000 years, which defines the Holocene (Committee on abrupt climate change 2002; Folland et al. 2001). This period coincides with the epoch of greatest human development and, for some authors, was a crucial factor in the onset and spread of agriculture and modern civilisation. However, even being of lower magnitude than the preceding ones, a detailed study of Holocene climate reveals important climatic variations (Fig. VI). Between 11,500-10,500 cal. BP climate was still cooler than present, while from 9,000 to 8,200 cal. BP it became wetter and warmer than present. This was followed by a short cold period of about 200 years, around 8,200 cal. BP (Alley et al. 1997). Between 8,000 and 4,500 cal. BP climate was somewhat warmer and wetter than today, reaching since then environmental conditions relatively similar to those found in present times, except for some cold phases (900-300 BCE, Iron Age Cold Epoch; 1,600-1800 CE, Little Ice Age) (Gribbin and Lamb 1978; Van-Geel et al. 1998) and warm episodes (900-1400 CE, Medieval Warm Period) (Bradley et al. 2001; Gribbin and Lamb 1978).
Nevertheless, the implications of these global climate reconstructions can be very variable at the regional or local scale. In particular, precipitation regimes show a considerably spatial heterogeneity in response to global climate phenomenons, as is the case for the North Atlantic Oscillation (NAO) or the ENSO (Muñoz-Díaz and Rodrigo 2003; Rodó et al. 1997). The heterogeneous response of precipitation against global climate changes is mostly due to the multiplicity of factors that determine precipitation: circulation of air masses, orography, ocean temperature. This heterogeneity is not only evident from current meteorological data, but also from palaeoenvironmental registers. Thus, for example, the cold period occurring 8,000 years ago reveals a wetter/colder climate in NE Iberian Peninsula (Pérez-Obiol and Julià 1994), but it was drier in the South of Spain (Magny et al. 2003), as well as in the Near East (Van Zeist and Bottema 1988; Willcox 1999; Willcox 2002).
At the regional scale, climatic reconstructions of high temporal resolution are already available for the Iberian Peninsula. Most of these studies are based on the reconstruction of past vegetation, either from palinology (Jalut et al. 2000; López-Sáez et al. 2003; Pérez-Obiol and Julià 1994) or from charcoal studies (Allué 2002; Badal 1990). The study of vegetation, however, has the disadvantage of reflecting both climate and human effects, which reduces its palaeoclimatic significance after the onset of agriculture (Lebreton et al. 2004). Another important source of information is the reconstruction of lake levels, based on a combination of stratigraphic and geochemical methods (Riera et al. 2004; Rodó et al. 2002), which has provided useful data on temperature and water availability. These studies are complemented by the diverse techniques used to reconstruct marine temperature, in the Atlantic Ocean and in the Mediterranean Sea, as sea temperature has a strong influence on continental climate (see e.g. Moreno et al. 2004; Nebout et al. 2002). For the last millennium, these data has been complemented by other approaches, such as dendrochronology (Creus and Saz 1999; Manrique-Menéndez and Fernández-Cancio 2000; Richter and Eckstein 1990) or documentary studies (Barriendos 1997; Barriendos and Martín-Vide 1998).