MARINE GEOLOGY AND GLACIOLOGY
|
|
Dynamic European Climate-VEGetation Impacts and Interactions - Data |
|
|
Contact: T. Giesecke tg@geus.dk |
PRIMARY DATA COLLECTION AND HANDLING
Palaeovegetation data.
Aim:
Organisation of fossil pollen and plant macrofossil data needed for improved palaeoclimate reconstruction.
Past vegetation cover and its dynamics are well documented in numerous well-dated pollen and plant macrofossil diagrams that have been published over the last 30 years. Many of these datasets are already organized in the
European Pollen Database (EPD)
(http://medias.obsmip.fr/paleo/epd/epd_main.html), which has been the official archive for European palynological data since 1989. The EPD contains an uneven distribution of over 900 sites which will be substantially improved in the frame of the present project. A developing independent plant macrofossil database will also be incorporated into the EPD.
Click on image to download in full size, JPG-file (330 KB)
Dating
Aim:
Development of age-depth models for each data site to improve data accessibility for data-model comparisons.
At present the EPD cannot be questioned to extract data for given time slices or periods as will be necessary for this project. Dating information exists for most sites, but consistent age-depth chronologies need to be developed in calendar years for data-model comparison. These chronologies will then improve considerably the value of EPD to the broader research community.
Modern Pollen Data
Aim:
To calibrate pollen with modern climate by compilation of surface sample data set and training of neural networks with modern bioclimatic variables.
Existing surface sample pollen data sets, which have been prepared in the course of numerous earlier projects, will be located and collated. If there are large geographical areas from which such data are lacking it may be necessary to collect and analyse additional samples. A more restricted amount of annual pollen deposition data exists through the
Pollen Monitoring Programme (PMP)
(http://www.ncdc.noaa.gov/paleo/pmp) and is compiled in PMPdata, a database that is also compatible in format with the EPD. These data reflect how the actual quantity of pollen varies annually and so have a temporal resolution and quantification which cannot be obtained through the more extensive surface sample collection.
EXTRACTING CLIMATE DATA FROM PALAEOVEGETATION RECORDS
Direct Palaeoclimate Reconstruction
Aim:
Derive palaeoclimate from pollen data using multiple methods including training the modern pollen data with modern bioclimatic variables in neural networks and classification of the fossil pollen data to generate European bioclimatic variables for the past.
The modern pollen data will be trained in neural networks using present day bioclimatic variables. The key variables that can be trained in this way are growing degree days (GDD), precipitation and minimum winter temperature, which are available on a grid at 10' intervals for Europe. The fossil pollen data covering the time periods to be studied will be classified into short time series using this model to generate dynamic European distributions of the bioclimatic variables. The bioclimatic variables generated will be compared with those obtained from inverse modelling (see below) and those available as GCM output. A limited number of new GCM model runs will be made to increase the number of palaeoclimate reconstructions available for comparison.
Inverse Modelling.
Aim:
To optimise extraction of palaeoclimatic information from pollen data by modification of DVMs for inverse modelling using fossil pollen data to generate European bioclimatic variables for the past.
We will use inverse modelling to reconstruct the climate scenarios that best fit the pollen data. These climate scenarios have the shape of probability histograms. The study of the divergences between the vegetation models for each of the bioclimatic variables considered will be instructive to indicate the different sensitivities of the DVMs to each climatic variable. An important task will be adaptation of algorithms to work with time series rather than single time-slices.
For more information see:
Edited by:
Niels E. Poulsen, GEUS -
nep@geus.dk
