The improvement in resolution of climate models has always been mentioned as one of the most important factors when investigating past climatic conditions, especially in order to evaluate and compare the results against proxy data. Despite this, only a few studies have tried to directly estimate the possible advantages of highly resolved simulations for the study of past climate change.
Motivated by such considerations, in this paper we present a set of high-resolution simulations for different time slices of the mid-to-late Holocene performed over Europe using the state-of-the-art regional climate model COSMO-CLM.
After proposing and testing a model configuration suitable for paleoclimate applications, the aforementioned mid-to-late Holocene simulations are compared against a new pollen-based climate reconstruction data set, covering almost all of Europe, with two main objectives: testing the advantages of high-resolution simulations for paleoclimatic applications, and investigating the response of temperature to variations in the seasonal cycle of insolation during the mid-to-late Holocene. With the aim of giving physically plausible interpretations of the mismatches between model and reconstructions, possible uncertainties of the pollen-based reconstructions are taken into consideration.
Focusing our analysis on near-surface temperature, we can demonstrate that concrete advantages arise in the use of highly resolved data for the comparison against proxy-reconstructions and the investigation of past climate change.
Additionally, our results reinforce previous findings showing that summertime temperatures during the mid-to-late Holocene were driven mainly by changes in insolation and that the model is too sensitive to such changes over Southern Europe, resulting in drier and warmer conditions. However, in winter, the model does not correctly reproduce the same amplitude of changes evident in the reconstructions, even if it captures the main pattern of the pollen data set over most of the domain for the time periods under investigation. Through the analysis of variations in atmospheric circulation we suggest that, even though the wintertime discrepancies between the two data sets in some areas are most likely due to high pollen uncertainties, in general the model seems to underestimate the changes in the amplitude of the North Atlantic Oscillation, overestimating the contribution of secondary modes of variability.
Climate has always a direct effect on all living organisms, and
always will have an influence on human affairs
Many uncertainties still affect climate models, particularly regarding their
sensitivity to changes in the external forcings
An important case study is represented by the evolution of European climate
during the mid-to-late Holocene (from 6000 years ago to present day). The
large number of proxy data available and the particular configuration of the
Earth astronomical parameters make it a useful period for the evaluation of
the models' response to changes in insolation
Different studies have been conducted in order to understand the mechanisms
driving the seasonal behaviour of European surface variables during the
mid-to-late Holocene.
(Left) Anomalies of zonal mean insolation on top of the atmosphere
(TOA) between 6000 years BP and pre-industrial period (PI). (Right)
Mid-to-late Holocene trends of the anomalies, with respect to present-day
values, of December and June TOA incoming insolation, calculated, according
to
Bearing this in mind, in recent years the application of regional climate
models for paleoclimate studies has become more frequent. For example,
In several studies, regional simulations of European climate during different times of the mid-to-late Holocene have been performed
In this paper we employ for the first time a regional climate model, the
COSMO-CLM (CCLM), for the investigation of the main climatic changes
that characterized Europe during multiple time slices of the Mid-to-Late
Holocene, with three main objectives:
Propose and test a model configuration suitable for paleoclimate studies Investigate the possible added value of highly resolved simulations arising in the comparison against proxy-reconstructions Analyse proxy and model mismatches, providing plausible physical interpretations of the dynamical processes responsible for them
Our discussion is structured as follows: in Sect. 2 the employed methodology, including a brief description of the models and the proxy data sets, is presented. Results are illustrated and discussed in Sect. 3: first a validation of the data for present-day conditions is conducted in order to test the performances of the model with the changes necessary for paleoclimate applications; then the mid-to-late Holocene simulations are compared against pollen-based reconstructions, trying, in a first instance, to highlight the advantages of the performance of highly resolved simulations specifically for this case of study; finally, physically plausible interpretation of the mismatches between the CCLM results and the reconstructions are proposed; the results of other studies are additionally discussed.
In this work we perform a set of climate simulations, covering several time slices of mid-to-late Holocene, employing models at different resolution.
The modus operandi consists of three parts and is based on the so-called First a transient continuous simulation is performed with the coupled
atmosphere-ocean circulation model ECHO-G, composed by the ECHAM4 We then select seven different time slices, at a temporal distance of approximately 1000 years from
each other, from 6000 years ago down to the pre-industrial period, 200 years before present, in accordance
to the time slices for which the pollen reconstructions are available. For every time slice, a simulation
is conducted, for a 30-year period, with the atmosphere-only global circulation model ECHAM5 Finally the ECHAM 5 outputs are further downscaled with the regional climate model COSMO-CLM model
version 4.8 clm 19 at a horizontal resolution of 0.44 longitude
degrees, using 40 vertical levels. The CCLM model
is a non-hydrostatic RCM with rotated geographical coordinates and a terrain following height
coordinate
In a first step we want to test whether the RCM setup and the applied model's
code modifications, required for implementing values of GHGs and astronomical
forcings, are suitable for paleoclimate studies. In order to set the values
of astronomical parameters for the corresponding investigation periods, we
apply the routine of
The setup of the COSMO-CLM is based upon the work of
Orography map of the COSMO-CLM simulation domain in rotated coordinates.
COSMO-CLM Main model configuration parameters.
Winter (left) and summer (right) temperature cost function estimates for the CCLM and the ECHAM5 models compared to the Proxy reconstructions for each time slice of mid-to-late Holocene. Values closer to 0 indicate a better agreement with proxy reconstructions.
For the model validation for present climate, the E-OBS
Subsequently, the results of the mid-to-late Holocene simulations are
compared against the data set of
Analysis of
The choice of the data set of
As a first step a control simulation has been performed with present values
of orbital parameters and greenhouse gases (Sect.
The simulation covers a 10-year period, between 1991 and 2000. Even if the length of this simulation can be considered as “critical” for the model's validation, we want to acknowledge that, due to computational reasons, it was not possible to cover a longer period.
As Fig. 3 but for
In Figs.
In the first column of each panel, the climatology of the different data sets is shown: the model is able to correctly reproduce, within a certain degree of accuracy, the climatology of the observations for both temperature and precipitation in winter and in summer.
Biases of seasonal means of Evapotranspiration (left), Latent
(centre) and Sensible Heat (right) fluxes, between the CCLM simulations and
the GLDAS data set, calculated for the reference period 1991–2000. As in the
previous figures, the area with a point represent the grid cells where the
anomalies between the two data sets are not significant, according to a
Student's
In the right column of every panel, temperature and precipitation values from
the present-day control run are directly validated, through a Student's
Based on these considerations, we suggest that the model reproduces
anomalously warm and dry conditions over a wide part of Southern Europe and
the Mediterranean basin, during summer, as a consequence of a wrong
conversion of energy towards latent heat in these regions. This hypothesis is
supported by the heat fluxes and evapotranspiration maps
(Fig.
Nevertheless the performances of the model with the applied changes are in
good agreement with the results of other works focusing on the same region
In a successive step, we conduct a comparison of the three models at
different resolution in order to estimate possible advantages in the use of
highly resolved simulations for paleoclimate studies. According to
Aiming at investigating the value added by highly resolved simulations for the comparison of changes in near-surface temperatures against proxy-reconstructions, we follow a two-step approach:
Firstly, we conduct a qualitative analysis of the simulations performed with three models at
different resolution in order to detect visible differences in the reproduced signals. Secondly, we employ a quantitative approach in order to estimate the skills of the RCM, in
comparison to the driving GCM, in reproducing the same mid-to-late Holocene changes in temperature as derived from proxy-reconstructions.
As a benchmark for such comparison we use the pollen-based temperature
reconstructions of
In Fig.
Maps of the anomalies between 6000 BP and the preindustrial period of
Consequently, we focus further analysis on the comparison between the ECHAM5 and the CCLM results. In both seasons additional details are easily detectable in the CCLM pattern. The coastline is also better reproduced in this case, resulting in a better detailed representation of the land-sea contrast, a more precise reproduction of surface processes and, consequently, leading to more suitable information for possible comparison against proxy-data. Nonetheless, the CCLM shows better defined patterns as a consequence of higher resolution, being able to discriminate higher spatial variability.
On the basis of such analysis, in the successive step, we try to quantify how
better the CCLM reproduces the reconstructed temperatures in comparison to
the ECHAM5. For this purpose we use an approach similar to the one employed
by
The values of the cost function for the two models are provided in
Table
As we can notice, even if not particularly large differences are present, the
Cost Function computed for the CCLM is in almost all the cases smaller than
the one for ECHAM5. In particular the CCLM results are, in some cases, closer
by more than 10 % to the reconstructions. It is important to mention that
the scale of the pollen-based reconstructions, considered for our analysis,
is closer to the resolution of the ECHAM5 than of the CCLM. As suggested by
Finally we focus on the comparison between the CCLM results and the pollen-based reconstructions. After analysing the differences between the two data sets and their temporal evolution, we propose, by means of correlations with trends of insolation and changes in atmospherical circulation patterns, physically plausible interpretation of the evinced mismatches.
Left: maps of
Figures
In winter, generally colder conditions are reproduced by the model over northern continental Europe, with
slightly warm biases over most of the South (Fig.
In summer, instead, CCLM results present positive anomalies over most of the
domain, with particularly pronounced values (in some case larger than
In addition to the previous analyses, the maps of temperature temporal
evolution are presented in Fig.
As in Fig. 7 but for
Mid-to-late Holocene temporal Evolution of the anomalies, with respect to the pre-industrial period, of near-surface temperature winter (first row) and summer (second row) seasonal means, derived from the CCLM simulations (left) and the pollen-based reconstruction (right). The maps show the slopes of the linear trends calculated, for every grid box, taking into consideration the uncertainties associated to the two data sets, by means of a weighted least squares method. The area masked out in grey, are the area where the trends are not significant, according to a F-test at a significance level of 10 %.
From these maps we see that in winter, even if over part of Southern Europe
the two data sets present similar trends, their behaviour is different in the
North: CCLM results show no significant trend (Fig.
In our analysis we adopt the method of
Figures
The MSLP pattern explaining most of the variance in winter resembles the
NAO (Fig.
In summer, instead, the first CCA pair (Fig.
Canonical correlation pattern pairs of MSLP (left) and T2M (right)
in
As in Fig. 10 but for
Canonical score series of the first two pairs of canonical correlation patterns of, respectively, MSLP (left column) and 2 m temperature (right column) winter seasonal mean anomalies.
As in Fig. 12 but for summer.
In the second CCA pair, the pattern of the mean sea level pressure
(Fig.
Consequently, we suggest that in summer, during mid-to-late Holocene, the
changes in circulation alone would not have been enough to explain the
variations in surface temperature, as reconstructed from the proxies. While
over Northern Europe the relatively good agreement between the temperature of
the two data sets over part of the domain suggests that for this region the
insolation is probably the main driver of change; for Southern Europe,
however, the role of land-atmosphere coupling needs to be considered
According to
It is important to mention that the behaviour of mid-to-late Holocene's
summer temperature over Europe has been highly debated during recent years.
While a dipole behaviour between Southern and Northern Europe has been
suggested by several studies based on pollen analyses
The latest hypothesis should be taken into account for the comparison between
pollen-based reconstructions and model simulations. Nevertheless, additional
investigations have shown that, when directly compared to the pollen record,
the mid-Holocene vegetation simulated from the output of climate models is
way too dry over Southern Europe, with an expansion of Mediterranean and
steppe/desert vegetation and contraction in forest cover, a direct
consequence of simulated warmer conditions
Based on these considerations, recognizing the data set of
An important benchmark for the comparison of our results against other
modelling studies is represented by the outcomes of the PMIP3 experiment
In this work we performed for the first time a set of highly resolved climate simulations over Europe for different time-slices of mid-to-late Holocene, by means of the state-of-the-art regional climate model COSMO-CLM.
As a first step, using the CRU and the E-OBS observational data sets as
benchmarks, a model setup suitable for paleoclimate investigations has been
tested for the reference period 1991–2000. The results show that the RCM is
able to reproduce realistic climatology with respect to the observations. The
largest biases arise in summer over Southern Europe where the model
reproduces warmer and drier conditions (
Successively, the results of mid-to-late Holocene simulations have been compared against a new pollen-based climate reconstruction data set. Winter and summer seasonal means of near-surface temperature have been considered for our analysis.
To begin with, the possible advantages of higher resolution models for paleoclimate applications have been investigated. The RCM seems to better reproduce the signal of the climate-reconstruction when compared to the driving GCMs, with a more detailed reproduction of the coast-line and better defined patterns. Additionally, using a quantitative approach, we have demonstrated that the results of the RCM are closer to the values of the reconstructions in comparison to the driving GCM, in some cases by more than 10 %. Considering also the final user perspective, the evinced results gave us concrete reasons for choosing to conduct highly resolved simulations for this particular case study.
Finally, the CCLM results are used in order to investigate the response of the climate system to changes in the seasonal cycle of insolation, with the aim of proposing plausible physical interpretations of the mismatches arising in the comparison against the reconstructions.
The results show that, in winter, over Southern Europe temporal behaviour and spatial distribution of temperature in the two data sets are comparable. Conversely, the model tends to reproduce generally colder conditions over central and northern continental Europe. Through the analysis of atmospheric circulation patterns we argue that this bias is due to a different representation by the model of the expected changes in circulation, as a result of reduced influence of westerly winds and an increased importance of secondary modes of atmospheric variability. Additionally, larger differences are present in Northeastern Europe, likely related to high uncertainties of pollen data over this area. In summer, the simulated northern conditions agree well with the proxy data over part of the domain. Their behaviour seems to be a direct response to insolation changes. Conversely, while the model produces warmer summer conditions over Southern Europe at mid-Holocene, in comparison to pre-industrial times, again mainly due to insolation changes, the pollen data exhibit an opposite trend. According to the results of previous works and to the analysis of atmospheric dynamics, we suggest that this behaviour is mainly due to a higher partition of radiation towards latent heat, resulting in a cooling effect of the surface that the model is not able to reproduce due to deficiencies in the representation of soil-atmosphere heat fluxes over this area. Nonetheless, it is important to mention that the validity of reconstructions of European summer temperature over the Mediterranean region based on pollen data has been highly questioned in recent years. Even though several evidences confirm an increasing trend of temperature over the area from 6000 BP to present day, joint efforts from specialists of different disciplines are still required in order to further clarify possible uncertainties.
This paper sets the basis for further investigations: in particular a set of new simulations with improved radiation schemes, soil properties and land use, could lead to important contributions to climate modelling and, consequently, to the improvement of future climate change projections.
Pollen-based reconstructions used in this study were published in Mauri
et al. (2015) and are available at:
The authors are grateful to the two anonymous referees for their constructive comments that helped to considerably improve the manuscript. This paper was supported by the Cluster of Excellence “Topoi – The Formation and Transformation of Space and Knowledge in Ancient Civilizations”. The computational resources were made available by the German Climate Computing Center (DKRZ) and the Freie Universität Berlin (ZEDAT). We would like to thank Achille Mauri and Basil Davis for providing the pollen-based reconstructions and for their continuous support and constructive discussions. We would also like to express our sincere appreciation to Janina Körper for designing and conducting the ECHAM5 climate simulations. A particular acknowledgment goes to Edoardo Mazza for his continuous support and intellectual debate. We would also like to thank Ingo Kirchner, Bijan Fallah, Nico Becker, Alexander Walter and John Walter Acevedo Valencia for the fruitful and interesting discussions. Additionally, we are particularly grateful to Jacqueline Harvey for proofreading the manuscript. Edited by: H. Goosse Reviewed by: two anonymous referees