Docstoc

WP1

Document Sample
WP1 Powered By Docstoc
					WP1 Data Harmonisation and Consolidation
Lead Contractor: Alterra; contributions this period from Alterra, Hadley, LSCE and EFI

3.1 Objectives

The objectives of this workpackage is to provide each of the datasets required by WP2, WP3 and WP4:

      1.   Compile datasets from selected suitable sites for validation of the TEMs from the full list of those
           involved in the CARBOEUROPE cluster and related projects.
      2.   Existing datasets of historical land-use and nitrogen deposition for the period 1900-2000 will be
           formatted for use WP3.
      3.   Existing data from forest inventories and agricultural production will be used to prepare maps of
           the changes of European carbon stocks between 1960 and 1990.
      4.   Satellite datasets of landcover, fAPAR and stand density/canopy structure will be prepared for
           Europe, and other key forest areas.

These objectives are associated to the following deliverables foreseen for the month indicated:
D1.1 Biome-specific datasets to drive and validate TEMs. (month 9)
D1.2 Atmospheric CO2 dataset for use in nowcasting system (month 9)
D1.3 Land-use and nitrogen deposition historical datasets (1900-2000) (month 12)
D1.4 Datasets of changes in European land carbon (month 18)
D1.5 Datasets of fAPAR for Europe (month 18)

Deliverables 1.1, 1.2 and 1.3 (landuse part) have been completed though updates/extensions are foreseen.
See next section. Compilation of Nitrogen deposition data (D1.3) is currently not persued as at this moment
none of the models using halfhourly or hourly timesteps is able to use nitrogen deposition. Deliverable 1.4
and 1.5 are in progress.

3.2        Methodology and scientific achievements

           Biome-specific datasets to drive and validate TEMs
           (Alterra and LSCE)

The main objective this task is to provide harmonised and consolidated biome specific datasets to drive,
calibrate and validate the models in WP2. We started with sites with easily accessible, good quality data
that had been used with one of the models already. This dataset is now ready and is being used to carry out
the first model optimisations (see section on WP2). It will be extended with more years of data and
additional sites to cover all global PFT’s with as broad a spectrum of sites as possible, varying in soil types,
dominant plant species and geopgraphical location (within the limits of the climate zone).

Twenty site datasets have been prepared from publicly available data (Fluxnet, Carbodata). The length of
the datasets ranges from one growing season (Upad) tot seven years (Harvard Forest). All data sets are
from the Northern Hemisphere (Europe, US or Canada) and most are from forests (figure 3.1a). Forests
have also the longest running time series. The few datasets from non-forested flux sites all are from North-
America.

In Europe, human pressure on land has since long converted the most fertile soils into agricultural land or
pastures. Poorer, sandy soils were left for forests. The past strong focus of CARBOEUROPE to forested
sites is thus reflected in a high representation of sandy soils within the European sites and in the datasets as
a whole (Figure 3.1b).

Before the datasets were put together an inventarisation was made of model needs. A distinction was made
between indispensable (meteorological, vegetation or soil) data, other information that was not crucial but
could be used if available, and measurements that could be used to compare models output to for
optimisation and validation (Appendix I-B and I-C).
                                                                                     Flux sites per soil texture type

                                                                          7




                                                        number of sites
                                                                          6
                                                                          5
                                                                          4
                                                                          3
                                                                          2
                                                                          1
                                                                          0
                                                                              clay    clay    silty  silt   loam loamy sandy sandy sand
                                                                                     loam     clay loam           sand clay loam
                                                                                             loam                      loam

                                                                                                       Europe   USA

Figure 3.1: (left) Location of flux sites in current data set (for site names and details see Appendix I-A).
(right) Distribution of flux sites in current data set over soil texture types


All three the (half)hourly models need incoming shortwave radiation, average temperature, vapour pressure
deficit, air CO2 concentration and preciptation over each time interval. Apart from these, all have further
needs for meteorological input, covering the total set of standard meteorological measurements in
CARBOEUROPE. Net and longwave radiation are the only variables outside the standard
CARBOEUROPE measurements. Only these mandatory input data were gapfilled. Gapfilling was done
using meteorological data from NCDC (8000 stations worldwide) and ERA15 data following a procedure
developed by Nicholas Viovy (LSCE-partner). No attempt was made to estimate errors made by gapfilling,
but gapfilled meteorological data were flagged in the final dataset. All driving variables were given along
with estimated errors (see below).

Soil data needs are limited to soil type and texture, which is often more information than is available
already. Information about the vegetation (apart from the PFT class) is limited to LAI, which is needed and
provided on a seasonal basis. Optionally, the models can use further information about plant and soil
carbon pools.

All 3 flux models were developed for their output to be compared against NEE, latent heat flux and
sensible heat flux, and optionally ground heat flux. MOSES and ORCHIDEE can also use calculated
radiation terms or plant carbon pools and LAI for that purpose. BETHY and ORCHIDEE can also use litter
carbon. Variables that are used for model output comparison were not gapfilled, and all gapfilling that had
occurred by others was removed. They were given along with preliminary estimated errors.

For the optimisation methods used in WP2 it is imperative to give a-priori estimates of uncertainties of
driving and validation data. The have been provided based on the folowing table. A distinction must be
made between stochastic errors, that may (partially) cancel out when averaged overv longer that halfhourly
time scales, and systematic errors that do not diminish in averaging. Table 3.1 presents rough estimates of
potential uncertainty on eddy correlation flux measurements. This analysis is based upon various studies
showing sensitivity of fluxes to treatment and calculation method as well as variation of representativivity.
The estimates are assuming high ignorance of the user (modeller) about maintenance conditions, quality
control and field conditions at the data source. In almost all individula cases the uncertainties are likely to
be much smaller and only in some cases a few sources of error may be stronger.

All meteorologiscal and flux data were converted to NetCDF format. Version numbers of all datasets are
included. Dataset are available (CAMELS partners only for now) from the following ftp adress ftp.bgc-
jena.mpg.de
Table 3.1 Estimates of potential uncertainty (%) on eddy correlation flux measurements




                                                  Variation time scale

                                                                         NEE day
                                                                                   1
                                                                                       NEE night, u*>0.2 m s
                                                                                       u*<0.1 m s
                                                                                       NEE, systematic night
                                                                                                                 1
                                                                                                                     NEE night, u*<0.1 m s
                                                                                                                     ms
                                                                                                                     NEE night, 0.1<u*<0.2
                                                                                                                                               LE
                                                                                                                                                    LE at RH>95%
                                                                                                                                                                   H
                                                                                                                                                                       Additional error
                                                                                                                                                                       > 10 mm h
                                                                                                                                                                       flux during rain
                                                                                                                                                                       flux
                                                                                                                                                                       systematic error on all
                    Type and time of flux




                                                                                                                        -1




                                                                                                                                                                                  -1
                                                                                                  -1
Type and source of error




                                                                                                             -




                                                                                                                                           -
instrument accuracy sonic                     Hour                         1       1                 0           1                 1           1       1           1        0            0
instrument accuracy IRGA                      Hour                         1       1                 0           1                 1           1       1           0        0            0
location and footprint
                                              Day                        20 20                       0           20               20           30 30 20                     0            0
(Rehbmann et al 2003)
stochastic error in turbulence
                                              Hour                       20 20                       0           20               20           20 20 20                     0            0
(Finkelstein 2001)
Night-time 'losses' (Kruijt et al, 2003)      Hour                         0       0              100 200 100                                  0       0           0        0            0
Angle of attack (Gash et al, 2003;
                                              Day                          5       5                 0           5                 5           5       5           5        0            5
Van der Molen et al, 2003)
Webb (density) correction
(open path only, p.m.)
Routine cleaning instrument -
                                              2-week                       5       5                 0           5                 5           10 15 10                     0            0
maintenance
Calibration - routine checks                  2-week                       5       5                 0           5                 5           10      0           1        0            0
tube delay correction
                                              Hour                         4       10                0           10               10           10 20               0        0            0
(closed systems only, Kruijt et al, 2003)
Spikes - uncorrelated
                                              Hour                       10 10                       0           10               10           10 10 10                   100            0
(Kruijt et al, 2003)
low and high frequencies, rotation            Day-
                                                                         15 20                       0           20               20           10 40               5        0            0
(Kruijt et al, 2003)                          season
Total relative uncertainty                    Mix of
                                                                         35 38                    100 202 103 38 48 26                                                    100            5
(geometric mean)                              above!



         Atmospheric CO2 dataset
         (LSCE)

The CO2 dataset consists of two parts:
     A historical dataset enabling to run the models from 1800- present
     A contemporary dataset for use in nowcasting system

The historical CO2 forcing is provided (following Rayner and Trudinger, CSIRO) as annual mean
concentration taken from a spline fit to the ice core record from Law Dome Antarctica (Etheridge et al.
1996) and a combined atmospheric record from the South Pole and Mauna Loa records of the Scripps
Institute of Oceanography (Keeling et al. 1995). The atmospheric record is calculated as a weighted sum
(0.75<mlo + 0.25*spo) which is a good fit to the global mean marine boundary layer value from the
Globalview compilation when both records overlap. The Law Dome values have been adjusted to bring
both records onto the same calibration scale. The spline uses different weightings for different periods to
account for varying data density (see Enting 1987 for details of the smoothing spline algorithm).
Throughout it is arranged to preserve decadal-scale variability but suppress shorter-term variations.Before
1960 (in the period covered by the ice core measurements) it produces a 50% attenuation at frequencies
less than about 20 years, during the period of the atmospheric record this frequency cut-off drops to about
12 years. See figure 3.2
Figure 3.2 The historical CO2 forcing as annual mean concentration


The contemporary dataset is based on the AEROCARB project database. The objective of this database is
to organise the existing atmospheric CO2 observations made in Europe into one single network and to
document it. This will provide a single, coherent dataset that can be used in models to retrieve the European
CO2 fluxes. This database is a cooperative effort to unify the dataset of many european institutions involved
in high-quality atmospheric CO2 measurements. The AEROCARB database contains data from January
1997 to December 2001. The data can be recorded by site (& heights) and year.

Four (five) types of files are available on the AEROCARB Database:
               the CO2 hourly means
               the CO2 daily means
               the CO2 monthly means
               the smooth curves
               and Figures (time series, diurnal cycle and number of selected data)
The current release of AEROCARB is available (limited access) via http://www.aerocarb.cnrs-gif.fr/.


         Datasets of historical land-use
         (Hadley Center)

A spatially-explicit, global-scale dataset of crop and pasture area for each year from 1700-1900 was
constructed by merging, interpolating and reconciling two independent source datasets (Klein Goldewijk,
2001 – KG; Ramankutty and Foley 1999 - RF). These data may be used as input to Terrestrrial Ecosystem
Models which require anthropogenic disturbance as a forcing.

The construction of annual pasture fraction datasets based on KG required interpolation and a number of
assumptions regarding fractional coverage and the exact nature of grazing land. WRT the latter it is
important to know its history as that determines to a large extend the (soil) carbon pool: is it intensively
grazed land converted from either natural grassland or forest, or is it natural grassland.
In the absence of more precise information, datasets of the fractional coverage of pasture at years other than
1700, 1750, 1800, 1850, 1900, 1950, 1970 and 1990 were obtained by simple linear interpolation between
those dates.

In many grid squares it was necessary to deal with the dual presence of both pasture from KG and crops
from RF, as a result of which a grid square could contain a pasture fraction of 1.0 and also a non-zero crop
fraction, giving a total “disturbed fraction” greater than 1.0. The crop and/or pasture fractions are therefore
adjusted in order to avoid total disturbed fractions of more than 1.0. The exact procedures are given in
appendix II.




Figure 3.3 Distributions of total anthropogenically-disturbed fraction of land for 1750 and1900.


The final datasets provide fractional coverage of crops and pastures at 0.5˚ resolution, each year from 1700
to 1990. The distributions of total anthropogenically-disturbed fraction of land (ie: crop + pasture) for 1750
and1900 are shown in Figure 3.3.


                                                                 Datasets of changes in European land carbon.
                                                                 (Alterra and EFI)

The latest estimates on European land carbon sinks are published in Janssens et al. (2003) and Nabuurs et
al. (2003). These studies are based on (historic) forest inventory data, agricultural production data and
models to convert these data to estimates of carbon stocks and fluxes. Figure 3.4 shows the evolution of the
European forest sector carbon sink over the period 1950-1999. The increasing sink strength must be
attributed to an increase in net annual increment of the forest since the 1960's and a more or less stable
harvesting rate. The increase in increment is probably influenced by a complex of factors: changes in the
age class distribution of the forest, increased nitrogren deposition and changes in forest management.

                                                         0.20                                                                                  Figure 3.4. Sinks in the European forest
                                                         0.18                  Tree Biomass
                                                                                                                                               sector, from 1950 to 1999, partitioned
                                                                                                                                               by compartments (Nabuurs et al. 2003).
Components of the total forest sector sink (Pg C y -1)




                                                                               Coarse woody debris
                                                         0.16
                                                                               Forest floor
                                                         0.14                  Mineral soil

                                                                               Wood Products
                                                         0.12
                                                                               Total
                                                         0.10

                                                         0.08

                                                         0.06

                                                         0.04

                                                         0.02

                                                         0.00
                                                                 1950   1955      1960        1965   1970   1975   1980   1985   1990   1995
                                                         -0.02



Janssens et al. (2003) give an estimate for the agricultural sector, where they conclude that grasslands act as
a (small) net sink, whereas croplands are net sources. Biomass changes in these systems are neglible, it are
the soils that determine if the system is a source or a sink. For these systems, the inventory method will not
add much to compare to the TEMs, since inventory estimates are good especially for the biomass part.
Precision for sink or source estimates will depend on the accuracy of modelling landuse changes, and of the
biomass model (Janssens et al.).

We can conclude that for the inventory part within Camels the most important issues are the changes in
forest biomass carbon stocks, and accurate estimates of landuse and landuse changes. We are now looking
into the national historic inventories to estimate carbon stock changes at a national/regional scale, also for
validation of the TEMs. Furthermore, the historic landuse maps (D1.3) will be compared to historical
inventory statistics when available.


         Dataset of fAPAR for Europe, and other remote sensing products
         (JRC)

Good progress has been made on both the fAPAR retrievals for the flux sites (11 will shortly be available
through the CarboData website), and the global fAPAR dataset. Additional requirements for fAPAR
products should be; (i) to extend the site-level fAPAR data to each of the 20 CAMELS flux sites, and (ii) to
produce global fAPAR datasets for other years (to capture interannual variability).

3.3 Socio-economic relevance and policy implication

Most of the datasets produced here primarily target scientific use and specifications are tailored to the
needs and parameterisations used in the models of this project. Underlying data sets generally have a wider
use and are subject to intense analysis by many scientist world wide. Within the project data analysis will
address the optimal use of existing data sources and the latest models to produce operational estimates of
the European land carbon sink. We will also exploring also the proximate causes of these sink (i.e. the
relative contributions of CO2 fertilisation, nitrogen deposition, climate variability, land management and
land-use change.

As such the development of these data will also support EU countries in meeting their obligations under
Kyoto, more specifically with respect to the reporting of sources and sinks of CO2 in a “transparent and
verifiable manner”.

3.4 Discussion and conclusion

Ideally the biome specific datasets should cover all Plant Functional Types and should geographically
representative. However, he current distribution of flux sites shows that this may be possible within certain
limits only. Non-forested flux sites from other than North-America are scarce and all non-forested
ecosystems will be represented by fewer sites and shorter time series. Forests in the northern hemisphere
and particularly North-America and Europe are already very well represented, while Asia and the Southern
hemisphere are very much underrepresented. There is a lower density of flux sites there and data are less
available. An additional effort will be done to have data from these places. It may be necessary to have our
data not publicly available for some time if we want to use unpublished or otherwise vulnerable data.

Existing data from forest inventories and agricultural production can be used to prepare maps of the
changes of European carbon stocks between 1960 and 1990. Possibilities, for going back in time, ideally till
1900, will be limited to a few countries where such data are available. Reconstructions might be possible
from the oldest data containing specific information on age class distribution. Inventory based data on land
use change (i.e. forest area contraction or expansion) will have to be made consistent with the historic
landuse maps discussed above. Prior or parallel to the data efforts a methodology needs to be developed to
use them in the models, both qualitatively – i.e. when and where land cover conversions took place – and
quantitatively – what area is affected and more importantly the effects on carbon stocks.
3.5 Plan and objectives for the next period

The objectives of this workpackage for the next period is to extend as necessary some of the datasets
described before:

    1.   Compile datasets from additional suitable sites for validation of the TEMs concentrating on sites
         representing non-forested PFTs and on sites in the Souithern hemisphere. A-priori estimation of
         uncertainties in fluxes will further constrained using site specific information.
    2.   Existing data from forest inventories and agricultural production will be used to prepare maps of
         the changes of European carbon stocks between 1960 and 1990. Focus will be on national historic
         inventories going back as long as possible. The historic landuse maps (D1.3) will be compared to
         historical inventory statistics. Prior or parallel to that a methodology to ingest these data in the
         models will be developed.
    3.   Satellite datasets of landcover, fAPAR and stand density/canopy structure will be prepared for
         Europe, and other key forest areas, including site specific products.
                    Appendix I (WP1) Biome specific datasets

                    I-A Sites locations and years included in the currently available biome-specific dataset.

Sitename         Latitude      Longitude      Years        Plant Functional Type               Dominant species                 Soil type
Aberfeldy        56°36'24" N   003°47'49" E   1997-1998    Temperate Needle-leaved Evergreen   Picea sitchensis                 (Stagno)humic gley
Bondville        40°00'__" N   088°18'__" W   1997-1999    C4 crops                            Corn                             Silt loam
Bordeaux         44°05'__" N   000°46'00" W   197-1998     Temperate Needle-leaved Evergreen   Pinus pinaster                   Podsolic
Brasschaat       51°18'__" N   004°31'__" E   1996-1998    Temperate Needle-leaved Evergreen   Pinus sylvestris                 Sandy deposits on tertiary loam
                                                                                                                                clay layer (poorly drained)
Castelporziano   41°45'__" N   012°22'__" E   1997-1998    Temperate Broad-leaved Evergreen    Quercus ilex                     Sandy
Flakaliden       64°07'__" N   019°27'__" E   1996-1998    Boreal Needle-leaved Evergreen      Picea abies                      Shallow till
Gunnarsholt      63°50'__" N   020°13'__" W   1996-1998    Boreal Broad-leaved Deciduous       Populus trichocarpa              Andisols
Harvard          42°32'16" N   072°10'17" W   1992-1999    Temperate Broad-leaved Deciduous    Quercus rubra, Acer rubrum,      Well drained acidic sandy loam /
                                                                                               Betula lenta, Pinus strobus,     poorly drained peat in low areas
                                                                                               Tsuga canadensis

Hesse            48°40'__" N   007°04'__" E   1996-2000    Temperate Broad-leaved Deciduous    Fagus sylvatica                  Gleyic luvisol on sandstone or
                                                                                                                                loam
Hyytiala         61°51'__" N   024°17'__" E   1996-2000    Boreal Needle-leaved Evergreen      Pinus sylvestris                 Haplic podsol on till material
                                                                                                                                dominated by fine sand
Little Washita   34°57'__" N   097°59'__" W   1997-1998    C3 grass                            Schizachyrium scoparium          clay loam
Loobos           52°10'__" N   005°44'__" W   1996-2000    Temperate Needle-Leaved Evergreen   Pinus sylvestris                 Sandy soil, humuspodsol
Metolius         44°27'__" N   121°33'__" W   1996-1997    Temperate Needle-leaved Evergreen   Ponderosa pine                   Alfic vitrixerands
Sky_oak_old      33°22'__" N   116°37'__" W   1997-2000    Evergreen shrubs                    Adenostoma fasciculatum,
                                                                                               Adenostoma sparsifolium,
                                                                                               Ceanothus greggii
Soroe            55°29'13" N   011°38'45" E   1997-1999    Temperate Broad-leaved Deciduous    Fagus sylvatica                  Cambisol, mollisol
Tharandt         50°58'__" N   013°34'__" E   1996-2000    Temperate Needle-leaved Evergreen   Picea abies                      brown earth (rhyolith)
Upad             70°16'53" N   148°53'05" W   1994         Tundra vegetation                   Eriophorum angustifolium,        Pergelic-Cryaquolls (moist) -
                                                                                               Carex aquatilis, C. bigelowii    Histic Pergelic Cryaquepts (wet)
Vielsalm         50°18'__" N   006°00'__" E   1996-1998    Temperate Broad-leaved Deciduous    Fagus sylvatica                  Dystric cambisol
Walker Branch    35°57'32" N   084°17'15" W   1995-1998    Temperate Broad-leaved Deciduous    Quercus alba, Q. prinus,         Infertile cherty silt loam
                                                                                               Carya ovata, Acer rubrum,
                                                                                               Liriodendron tulipifera, Pinus
                                                                                               taeda
Weidenbrunnen    50°10'__" N   011°53'__" E   1996-1999    Temperate Needle-leaved Evergreen   Picea abies                      Acidic cambisol
I-B Driving data and site-specific input parameters of the CAMELS-models. Data are either mandatory (M)
for model runs or can be used optionally (O), and should be available on an hourly (h), daily (d) or seasonal
(s) basis.


                                            MOSES        ORCHIDEE        BETHY       LPJ        summary
Meteorological Data:
Sum of incoming shortwave radiation         M (h)        M (1/2h)        M (h, d)    M (d)      M(1/2 h, d)
PAR                                         O (h)                        M (h, d)               M (h, d)
Sum of net radiation                        M (h)                        M (h, d)               M (h, d)
Long-wave radiation down                    M (h)        M (1/2h)                               M (1/2 h)
Air temp av.                                M (h)        M (1/2h)        M (h)       M (d)      M (1/2 h, d)
Air temp max.                                                            M (d)                  M (d)
Air temp min.                                                            M (d)                  M (d)
Soil temp. av.                              O (h)                                               O (h)
VPD av.                                     M (h)        M (1/2h)        M (h)                  M (1/2h)
CO2 av.                                     M (h)        M (1/2h)        M (h)                  M (1/2h)
Sum of precipitation                        M (h)        M (1/2h)        M (d)       M (d)      M (1/2h, d)
Soil water content av.                      O (d)                        M/O (d)                O (d)
Wind speed av.                              M (h)        M (1/2h)        O (h)                  M (1/2h)
Air pressure                                O (h)        M (1/2h)        O (h)                  M (1/2h)
Met. Data for Model spinup                  O (h)        O (h)           O (d, m)    O (m)      O (h, d, m)
Soil:
Soil type                                   M            M                                      M
Soil color dry - for albedo                              O               O                      O
Soil color wet - for albedo                              O               O                      O
Soil texture: sand, silt, clay              M            M               M           M          M
Soil depth                                  O            O               M           M          M
SOC                                         O (M)        O                                      O
Vegetation:
Vegetation Type                             M            M               M           M          M
Fraction of each veg. Type                  M            M               M           M          M
Dominant Species                            M                                                   M
Other Species                               M                                                   M
Leaf Area Index                             M (s)                        M (s)       M (s)      M (s)
Plant C Pool – herbaceous (leaves...)       O (m)                                    O (m)      O (m)
Plant C Pool -- woody                       O (m)                                    O (m)      O (m)
Plant C Pool – coarse roots                 O (m)                                    O (m)      O (m)
Plant C Pool – fine roots                   O (m)                                    O (m)      O (m)
Litter C Pool                               O (m)                        O (m)       O (m)      O (m)
N-Deposition                                                                         O (m)      O (m)
Plant N concentration-leaves                                             O                      O
I-C Variables to be used for optimisation and validation of CAMELS-models. Data are either mandatory
(M) for model runs or can be used optionally (O) on an hourly (h), daily (d) or seasonal (s) basis.

                                  MOSES        ORCHIDEE       BETHY      LPJ       summary

NEE                               M (h)        M (1/2h)       M (h)      M (d)     M (1/2h)
Sum of latent heat flux           M (h)        M (1/2h)       M (h)                M (1/2h)
Sum of sensible heat flux         M (h)        M (1/2h)       M (h)                M (1/2h)
Sum of soil heat flux             O (h)                       O (h)                O (h)
Net Radiation short wave          O (h)        O (1/2h)                            O (1/2h)
Net Radiation long wave           O (h)        O (1/2h)                            O (1/2h)
Plant C Pool – herbaceous
(leaves...)                       O (m)        O (m)                     O (m)     O (m)
Plant C Pool -- woody             O (m)        O (m)                     O (m)     O (m)
Plant C Pool – coarse roots       O (m)        O (m)                     O (m)     O (m)
Plant C Pool – fine roots         O (m)        O (m)                     O (m)     O (m)
Litter C Pool                                  O (m)          M (m)      O (m)     O (m)
LAI                               O (w)        O (w)                               O (w)
Appendix II Processing of land use data

The RF data on crop cover fractions were already in a form suitable for use in this work, so required no
initial processing. However, the construction of annual pasture fraction datasets based on KG required
interpolation and a number of assumptions regarding fractional coverage and the exact nature of grazing
land. Combination of RF crop fractions and KG-based pasture fractions required further processing to
reconcile inconsistencies between the two datasets, ie: crop fractions and pasture fractions which summed
to more than 1.0 in a grid square.

Distinguishing pasture from natural grazing

KG describes global land cover with the following classes: Cultivated land, Pasture/land used for grazing,
Ice, Tundra, Wooded tundra, Boreal forest, Cool conifer forest, Temperate mixed forest, Temperate
deciduous forest, Warm mixed forest, Grassland/Steppe, Hot desert, Scrubland, Savanna, Tropical
woodland, Tropical forest

The KG class “pasture/land used for grazing” includes both (i) intensively grazed land which was either
natural grassland or forest cleared for livestock grazing (“pasture” as defined here) and also (ii) uncleared
land which is grazed, which also is/would have been grazed by wild fauna. To distinguish the two types of
grazing land and hence avoid specifying uncleared land as cleared land, reference was made to the
reconstruction of potential vegetation also by RF. In grid squares specified as “pasture/land used for
grazing” by KG, the type of grazing was identified as either “pasture” or “natural” according to the
potential vegetation class allocated to that grid square by RF as shown in Table 1.

Table 1. Use of potential vegetation to distinguish pasture from natural grazing
Potential Vegetation (Ramakutty and Foley 1999)                                Type of grazing land
Savannah                                                                       Natural grazing
Dense shrubland
Open shrubland
Tundra
Desert
Polar desert
Tropical Evergreen Forest/Woodland                                             Pasture
Tropical Deciduous Forest/Woodland
Temperate Broadleaf Evergreen Forest/Woodland
Temperate Needleleaf Evergreen Forest/Woodland
Temperate Deciduous Forest/Woodland
Boreal Evergreen Forest/Woodland
Boreal Deciduous Forest/Woodland
Evergreen/Deciduous Mixed Forest/Woodland
Grassland/Steppe

The distribution of “pasture/land used for grazing” in the KG timeslice years was therefore used in
conjunction with the potential vegetation class distributions to create datasets identifying pasture as the
dominant land cover class at these dates.

Inference of fractional coverage of pasture

The presence of pasture in a 0.5˚grid square implies that this is the dominant land cover class, but does not
necessarily mean that the fractional coverage of pasture is 1.0. Similarly, the absence of pasture in KG
does not imply a true total absence of pasture, merely that pasture is not the dominant land cover class.
However, there is no more precise information on the actual fractional coverage of pasture. Therefore it
was assumed that the presence or absence of pasture in KG implied pasture fractions of 1.0 or 0.0
respectively. Although the assumption of total coverage may lead to an overestimation of the real pasture
fraction in grid squares where pasture is present in KG, at large-scale averages this is likely to be
compensated by underestimations in grid squares where pasture is not present in KG. Indeed, the
distributions of dominant classes by KG at 0.5˚resolution were specified in order to give realistic regional
totals (Klein Goldewijk, personal communication).

Interpolation of pasture data

In the absence of more precise information, datasets of the fractional coverage of pasture at years other than
1700, 1750, 1800, 1850, 1900, 1950, 1970 and 1990 were obtained by simple linear interpolation between
those dates.

Reconciling inconsistencies between the crop and pasture.

In many grid squares it was necessary to deal with the dual presence of both pasture from KG and crops
from RF. Considering first the KG timeslice years, a grid square could contain a pasture fraction of 1.0
and also a non-zero crop fraction, giving a total “disturbed fraction” greater than 1.0. This should then
require an adjustment of either the crop fraction or the pasture fraction.

However, if the RF crop fraction is less than 0.5 then this may not actually be an inconsistency as pasture
may be the dominant class in terms of covering more of the grid square. Pasture fractions as small as 0.5
could therefore be considered to still reflect the specification of pasture as the dominant class by KG.

Extending this argument to years between timeslice years, pasture fractions up to half of those specified in
the interpolated dataset could still be considered to be consistent with KG. Therefore, a further dataset was
constructed to represent the minimum pasture fractions consistent with KG. At timeslice years, the
minimum pasture fraction was 0.5 in grid squares where KG specifies pasture as the dominant class. In
intermediate years, the minimum pasture fraction was obtained by linear interpolation between the
minimum pasture fractions of the timeslice years.

The crop and/or pasture fractions are therefore adjusted as shown in the following figure, in order to avoid
total disturbed fractions of more than 1.0.

                                                      Figure. Procedure for adjusting crop and pasture
                                                      fractions if inconsistent. Minimum pasture = 0.5 ×
          Crop +
                            no     No adjustment to   pasture
          pasture > 1.0?
                                   pasture or crop



                   yes


          Crop +
                            no     Pasture = 1.0 -
          minimum
                                   crop
          pasture > 1.0?


                   yes

         Crop = pasture =
         0.5

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:8/19/2011
language:English
pages:14