Validation of chironomid-inferred temperature reconstructions in Iceland : the potential for reconstructing quantitative changes in Holocene climate

Validation of chironomid-inferred temperature reconstructions in Iceland: the potential for reconstructing quantitative changes in Holocene climate Subfossü chironomids in short cores from two lakes in western Iceland were analysed and a chironomidmean luly air temperature transfer function applied to produce chironomid-inferred temperature reconstructions for the recent past. These reconstructions were compared with local meteorological data in order to evaluate the technique in Iceland. The chironomidinferred temperature reconstructions showed similar patterns and magmtudes of change to those recorded in the instrumental data, but the chironomid temperature reconstructions slightly underpredicted the observed temperatures. This suggests that the chironomids are able to reconstruct sequence trends of temperature change although the accuracy of the actual figures produced needs to be addressed. Subfossü chironomids in Iceland can be used to reconstruct the relatively small magnitude temperature changes of the last few hundred years, and also the larger magnitude changes expenenced earher in the Holocene.


Introduction
Concern over the extent to which anthropogenic activity has forced recent changes in climate and could contribute to future climatic change has been the focus of many recent studies (Intergovernmental Panel on Climate Change, IPCC, 2001. To try to evaluate this it is essential to understand past natural climatic variability Instrumental meteorological records are only available for c.150 years so it is necessary to obtain proxy climate data for periods prior to this in order to validate climate modeis that predict climatic changes. The study of biological proxy data also has the potential to provide information on the possible ecologjcal effects any future climatic changes may have on lake ecosystems and their biota (e.g. Quinlan et al. 2005).
Chironomids (Insecta; Diptera; Chironomidae) are a biological proxy frequently used in palaeoenvironmental and palaeoecologjcal studies. The chitinous head capsules (see Figure 1) of chironomid larvae preserve in lake Sediments and can be isolated from the Sediment and analysed to provide information about past environmental conditions within the lake and local environment (Porinchu & MacDonald 2003). The potential of subfossil chironomids to reconstruct past changes in temperature has been demonstrated by a number of Lateglacial studies undertaken in the northern hemisphere (e.g. Bedford et al. 2004;Brooks & Birks 2000a, b;Porinchu et al. 2003). Recently, chironomid studies have been used to reconstruct Holocene palaeoclimatic changes (e.g. Kurek et al. 2004;Langdon et al. 2004;Velle et al. 2005a); however, the magnitude of temperature change in the Holocene is much smaller than over the last Glacial -Interglacial Transition, and much debate surrounds the use of subfossil chironomids to produce reliable temperature reconstructions for this period (e.g. Larocque & Hall 2003;Velle et al. 2005b). Caseldine et al. (2003) produced the first quantitative palaeolimnologjcal study from Iceland, using the Norwegjan chironomid-mean July air temperature transfer function (Brooks & Birks 2001;unpub.) showing the potential of subfossil chironomids present in Icelandic lakes as a climate proxy. The recent development of a chironomid-inferred mean July air temperature calibration model for Iceland (Caseldine et al. 2006; provided the opportunity to try to evaluate and validate the technique within Iceland. 2.1 Site selection and fleldwork Cores were obtained from two lakes situated in western Iceland, Baulärvallavatn and Saurarvatn, near to the meteorological Station at Stykkishölmur which has the longest instrumental temperature record for Iceland ( Figure 2). Baulärvallavatn is a relatively deep lake (Z 46 m) situated at 193 m above sea level, v max ' ' while Saurarvatn has a maximum depth of 14.3 m and is located close to sea level. The study sites were initially selected and sampled during the development of the Icelandic chironomid training set (Langdon et al.). Detailed bathymetries were produced and short cores obtained from each lake ( Figure 3). Cores were taken using a Renberg corer (Renberg 1991) and subsampled in the Seid. At Baulärvallavatn the cores were taken from a relatively Hat bottomed part of the lake at c.20 m depth as the lake bottom shelved quite steeply in deeper areas. Other data from Baulärvallavatn suggest that the depth of sampling location does not iniluence the chironomid data obtained (Holmes 2006). At Saurarvatn, the core was obtained from the largest of the three basins in the lake. Information about the physical characteristics of the lakes is detailed in Table 1.

Laboratory methods
Sedimentological and dating analyses. A subsample from each sample was analysed for CN (total %C, total %N and CN) using a Carlo Erba Elemental Analyser model NA2500. Particle size was analysed using a Saturn DigiSizer, and magnetic susceptibility measured using a Bartington MS2 Susceptibility System. Radioisotopic analyses were carried out at the University of Exeter and National Oceanography Centre, Southampton. Samples were analysed for 210Pb, 214Pb (Southampton), 226Ra (Exeter), and 137Cs by gamma spectrometry in order to derive chronological information about the lake Sediments (Appleby 2001). Chironomid analysis. Samples were prepared for chironomid analysis using Standard techniques with ultrasound treatment used to clean the head capsules in order to aid identification (Lang et al. 2003 Langdon et al.) to each sample using C2 (Juggins 2003).
3 Results 3.1 Core chronologies 137Cs and 210Pb provided chronologies for the cores from Baulärvallavatn and Saurarvatn. The core from Baulärvallavatn Covers a period of about 130 years (back to c.1870 AD). A peak in 137Cs at 9.25 cm represents the 1963/64 fallout maximum, while the 210Pb data suggest Sedimentation rates of 1-2.5 mm year1 (4-10 years cm1), with higher Sedimentation rates at the top of the core, perhaps due to a relative lack of compaction at the top of the core.There is close agreement between the 137Cs and 210Pb ages which suggests the 210Pb chronology is accurate.The core from Saurarvatn is thought to represent the period c.1950-2003 with a clearly defined 137Cs peak representing the 1963/64 weapon's fallout maximum at a depth of 20.5 cm. The 210Pb data suggest a linear Sedimentation rate of 0.2 mm year4 (5 years cm1) between 14-23 cm and that the upper 14 cm Covers the period 1995-2003, with a very high average accumulation rate of 1.75 cm year4 in this period, probably due to increased inputs of Sediment from the lake catchment.

Sedimentological analyses
Values of both %C and %N ( Figure 4a) are relatively low in Baulärvallavatn, reflecting the low productivity of the lake, and allochthonous origin of most of the Sediment within the lake. Increases in both elements in the upper Sediments are thought to be due to the greater proportion of undecomposed organic matter found here, but possibly reflect nutrient increases within the lake/catchment. The levels of %N and %C (Figure 4b) are higher in Saurarvatn reflecting its slightly more productive nature. %N shows an increasing trend up the core, while %C values remain fairly constant. Variability in both elements near the top of the core is thought to reflect inwash from the catchment which is responsible for the higher Sedimentation rates in this part of the core.

Chironomid assemblages
At Baulärvallavatn the chironomid stratigraphy ( Figure 4a) is dominated by Heterotrissocladius grimshawi-type, a cold stenotherm indicative of oligotrophic lakes, with values of between 40-67%. A number of rheophilic taxa, such as Eukiefferiella spp. and Diamesinae, are present, which along with the Simuliidae suggest some riverine influence on the lake. At Saurarvatn Psectrocladius sordidellus-type, Chironomus anthracinus-type, and Heterotrissocladius grimshawi-type are the most common taxa present throughout the core ( Figure 4b). Taxa, such as Chironomus anthracinus-type,, Cricotopus sylvestris-type and Orthocladius oliveri-type, which are thought to live in more productive ecosystems where macro-  (Caseldine et al. 2003), Vatnamyri and Hämundarstaöahäls (Caseldine et al. 2006), Torfadalsvatn, Viöarvatn (inset map) (Axford et al. 2007 3.4 Chironomid-inferred temperatures Baulärvallavatn. The chironomid-inferred temperature (C-IT) reconstruction from Baulärvallavatn shows little Variation through the core, with low temperature variability (0.78°C). No changes in temperature are greater than the sample specific prediction errors (SSPEs) and so are not significant statistically. The overall pattern is a slight warming trend. The C-IT reconstruction is compared to the instrumental meteorological temperature record from Stykkishölmur ( Figure 5a). The Stykkishölmur temperature data have been adjusted to the altitude of Baulärvallavatn (c.200 m) by applying a lapse rate of 0.65°C 100 m1.
Saurarvatn. The ränge of reconstructed temperatures is 1.1°C , smaller than the mean SSPE of 1.12°C, and as a result, none of the temperature changes are significant statistically. However, when the C-IT reconstruction is plotted alongside the temperature data from Stykkishölmur (Figure 5b) it can be seen that there are many similarities between the two records. Again the C-ITs consistently underpredict the meteorological data, although the predicted values (including SSPEs), in most cases, overlap with the meteorological data values. Both records show cooling during the 1950s-1970s followed by a period of warming in the early-1980s. A temperature decrease from the late-1980s to the early-1990s is also evident in both records, and the late-1990s to early 2000s warming trend seen in the chironomid data matches very well with the instrumental data in terms of pattern and magnitude of changes.

Discussion
The comparison of the C-IT reconstructions with the Stykkishölmur meteorological data suggests that the subfossil chironomids from Iceland do respond to, and are able to successfully reconstruct the relatively small magnitude variations in temperature that have occurred during the recent period, despite the observed temperature changes in the meteorological record being close to the error limits of the chironomid-mean July air temperature calibration model. However, at present, the reconstructions produced for the two sites studied here are underpredicting actual values, although at Baulärvallavatn all the reconstructed temperatures and their SSPEs are deemed similar to the running mean of the meteorological data. 4.1 Factors influencing the midges Since early work on quantitative C-IT reconstructions the question of whether midges are really responding to climate, either directly or indirectly, has been raised (Hann et al. 1992 & Hann 1987). The data presented in this paper do suggest the chironomids are responding to and therefore reconstructing changes in past temperatures, however, as is the case in the majority of chironomid studies, multiple factors could be influ- encing the midges, and as a result a multi-proxy study should be used where possible (Battarbee 2000). For both sites the patterns of the C-IT reconstructions are very similar to those in the meteorological data. However, over the short timescales covered by this study, it could be possible that changes in nutrient Status within the lake are greater than the changes in climate, and therefore that the chironomids are responding to the changing nutrient levels (Dalton et al. 2005 4.2 Underprediction of temperatures by the chironomid data The C-IT reconstructions underpredict the temperatures as recorded by the meteorological data. There are a number of possible reasons for this. Baulärvallavatn and Saurarvatn occur near the upper end of the temperature gradient of the Icelandic training set (Langdon et al), and many transfer functions underpredict temperatures at the upper limit of the gradient covered (Brooks & Birks 2000b). This can be overcome by increasing the number of sites in the training set and the length of gradient these sites cover; this would also act to lower the error terms of the model. It is also possible that the nature of the climatic data used caused the underprediction. The climate data used in the calibration model is from the period 1961-1990. Post-1980 temperatures have been increasing, and using climatic data from Stykkishölmur, it is thought this could account for c.0.5°C (nearly 50%) of the underprediction (Holmes 2006). Dalton et al. (2005) found that C-ITs for a Scottish loch were lower than those inferred from Vegetation Reconstruction des temperatures moyennes de l'air au mois de juillet deduites par les chironomides relativement aux donnees de temperature de juillet de Stykkishölmur. Les barres d'erreur des valeurs relatives aux temperatures deduites par les chironomides indiquent les SSPE (Sample Specific Prediction Errors). Un taux de 0,65°C 100 m1 a ete applique aux donnees meteorologiques de maniere ä prendre en compte la difference de 200 m environ entre Baulärvallavatn et Stykkishölmur. a) Baulärvallavatn, b) Saurarvatn reconstructions and noted that snowbeds are present in the catchment for much of the year, suggesting that summer snowmelt may keep the loch water cooler during the summer. This would lead to the presence of chironomid assemblages representative of areas with cooler mean July air temperatures (Dalton et al. 2005). At Baulärvallavatn, snowbeds are present within the catchment and melt from these might cause a slight reduction in the temperature of the lake water in relation to the air temperature. Saurarvatn, located near to sea level, is not influenced by snowbeds, so this would not account for the underprediction at this site.
Such underprediction was also seen by Granados & Toro (2000) who found that their chironomid temperature reconstruction showed the same general trends as local meteorological data, although the actual values produced were not reliable. Heiri et al. (2003) found that the most recent samples in a core from Hintburgsee produced C-ITs that were too high, possibly due to anthropogenic impacts on the lake ecosystem; this is thought unlikely at the Icelandic sites. In contrast, a number of studies have found very little discrepancy between modern observed temperatures and C-ITs. Apart from the study by Granados & Toro (2000), the only other example of quantitative chironomid records being compared to meteorological records is the study by Larocque & Hall (2003) in which C-TT reconstructions from four Swedish lakes were compared to local instrumental climate data. In their study, the majority of the C-ITs were considered to reconstruct relatively accurately, as in most cases the instrumental data feil within the SSPEs of the reconstructed temperatures. Brooks & Birks (2001) reconstructed the modern tem-perature of Loch an Uaine at 10.5°C while instrumental data recorded a temperature of 10.7°C and Langdon et al. (2004) produced a reconstructed temperature of 14.6°C for Talkin Tarn which compared well with the present mean July air temperature of 14.8°C.

Holocene timescale studies
The data presented above suggest that the subfossil chironomids produce useful temperature reconstructions during the instrumental period. It is important to know whether this is also the case over Holocene and longer timescales, when the need to reconstruct climate is greater as there are no instrumental records available. It is possible that factors influencing the midges have changed through the Holocene (Larocque & Hall 2003;Velle et al. 2005a), and it is therefore necessary to carry out some kind of Validation of the C-IT reconstructions produced for the Holocene. This can be done by comparison with other terrestrial data, and also with marine and ice core data.
Chironomids from two sites in the north of Iceland, Hämundarstaöahäls and Vatnamyri (see Figure 2) were analysed and the C-IT reconstructions produced were remarkably similar both in terms of patterns and values, even though the actual chironomid assemblages present in the two sites did differ (Caseldine et al. 2006). When the chironomid data are compared with offshore records it is also clear that the patterns and magnitude of changes are very similar in the terrestrial and marine proxy records (Caseldine et al. 2006). Axford et al. (2007) produced C-IT reconstructions from three Sites in northern Iceland (Figure 2). The reconstructions produced similar trends although there were some differences, particularly in the late Holocene (Axford et al. 2007). These data were compared with that from Efstadalsvatn, NW Iceland ( Figure 2) and a similar pattern of Holocene temperature development was seen (Axford et al. 2007). The fact that a number of temperature reconstructions from lakes located in different areas of Iceland are producing similar trends and magnitudes of changes in temperature suggests that the changes in chironomid assemblages are being driven by regional changes in temperature rather than by other, site specific, factors such as lake ontogeny and catchment development.
Although the validity of the Holocene sequences is supported it is difficult to evaluate the actual figures produced. New high resolution reconstructions of sea surface temperatures (SSTs) (e.g. Bendle & Rosell-Mele 2007) will provide further opportumties to evaluate Holocene timescale chironomid temperature reconstructions. SST reconstructions would be expected to show similar patterns of Holocene climate development to the chironomid sites studied, due to the near-coastal location of these sites.

Conclusion
These results suggest that subfossü chironomids can provide a reliable estimate of July air temperatures, especially temperature trends, even though the error limits of the chironomid-inferred mean July air temperature calibration model overlap with the magnitude of recent Holocene temperature changes. It is possible the error terms will be reduced by the future expansion of the training set. The chironomids seem capable of reconstructing both small magnitude temperature changes as expenenced during recent times, and larger magnitude changes as experienced earlier in the Holocene. They are however best used as part of a multi-proxy, multi-site study where other proxies are able to support the relatively subtle changes inferred based on analysis of the subfossil chironomids. -In: Journal of palaeolimnology 23:77-89. mean luly air temperature transfer function applied to produce chironomid-inferred temperature reconstructions for the recent past. These reconstructions were compared with local meteorological data in order to evaluate the technique in Iceland. The chironomidinferred temperature reconstructions showed similar patterns and magmtudes of change to those recorded in the instrumental data, but the chironomid temperature reconstructions slightly underpredicted the observed temperatures. This suggests that the chironomids are able to reconstruct sequence trends of temperature change although the accuracy of the actual figures produced needs to be addressed. Subfossü chironomids in Iceland can be used to reconstruct the relatively small magnitude temperature changes of the last few hundred years, and also the larger magnitude changes expenenced earher in the Holocene.