Map of the Arctic (1595) by Mercator

W.P 1.4: Land cover and Fire and their representation in models.

Objective 1: To amalgamate high latitude land cover maps derivedfrom remote sensing, assess their consistency and uncertainty, andtransform them into forms suitable for carbon, water and climate models.

Objective 2: To amalgamate the different sources of information aboutfire at high latitudes derived from remote sensing, assess their consistencyand whether they reveal any trends, and define methods to interface thesedata with carbon and climate models.

1) MODIS Product MCD12C1 (MODIS LC)Source: MODIS sensor on-board the Terra and Aqua satellites.Classes:17 classes according to the International Geosphere Biosphere Programme (IGBP).Spatial Resolution: Global coverage at 0.05 degreesTemporal Resolution: Monthly, 2001-2007.Available Format: HDF-EOS. Obtained From: https://lpdaac.usgs.gov/MONARCH Availability: Global coverage at 0.5 degrees for year 2001-2007 in NetCDF format.

2) Global Land Cover 2000 (GLC2000)Source: VEGETATION sensor on-board the SPOT-4 satellite.Classes: 23 classes according to the Land Cover Classification System (LCCS) by the Food and Agriculture Organization (FAO).Spatial Resolution: Global coverage at 0.009 degrees (~1 km at equator).Temporal Resolution: Annually, 2000.Available Format: Binary/TIFF/ESRI/IMG.Obtained From: http://bioval.jrc.ec.europa.eu/products/glc2000/glc2000.phpMONARCH Availability: Global Coverage at 0.5 degrees for year 2000 in NetCDF format.

3) GlobCoverSource: MERIS sensor on-board the ENVISAT satellite. Classes: 23 classes based on the Land Cover Classification System (LCCS) by the Food and Agriculture Organization (FAO).Spatial Resolution: Global coverage at 300 m.Temporal Resolution: Annually and bimonthly, 2005 & 2009.Available Format: GeoTIFF.Obtained From: http://ionia1.esrin.esa.int/MONARCH Availability: Global coverage at 0.5 degrees for years 2005 and 2009 in Netcdf format.

4) MODIS Product MOD44B (MODIS VCF)Source: MODIS sensor on-board the Terra and Aqua satellites.Classes: 3 classes, tree, herbaceous vegetation and bare ground coverage.Spatial Resolution: Global coverage at ~0.0045 degreesTemporal Resolution: Annually, 2001 only.Available Format: TIFF. Transformed into NetCDF format for the MONARCH project.Obtained From: http://www.landcover.org/data/vcf/MONARCH Availability: Global Coverage at 0.50 degrees for year 2001 in NetCDF format.

Individual classes of MODIS,GLC2000 andGlobCover were grouped into 3 broad Classes:Tree, herbaceous and bare ground coverage.e.g. GlobCover: Sparse(<15%) Vegetation:0% Tree, 15% Herbaceous and 85% Bare Ground.

•Methodology is to some extend subjective.•MODIS VCF showed significantly more herbaceous coverage in both regions.•Despite relatively straight-forward conversion even Bare Ground coverage demonstrated significant deviations,especially for GlobCover.•LPJs naturally driven land cover is also shown.

Fig 1: Area in N. America assigned to tree, herbaceous and bareground coverage for the 4 land cover data sets and the LPJ model.

Fig 2: Area in Eurasia assigned to tree, herbaceous and bareground coverage for the 4 land cover data sets and the LPJ model.

Fig 3: Maps of tree coverage in N.America for the 4 land cover data sets and LPJ.

Fig 4: Maps of herbaceous coverage in N.America for the 4 land cover data sets and LPJ.

Fig 5: Maps of bare ground coverage in N.America for the 4 land cover data sets and LPJ.

Fig 6: Maps of tree coverage in Eurasia for the 4 land cover data sets and LPJ.

Fig 7: Maps of herbaceous coverage in Eurasia for the 4 land cover data sets and LPJ.

Fig 8: Maps of bare ground coverage in Eurasia for the 4 land cover data sets and LPJ.

SDGVM was driven using GLC2000,MODIS and MODIS VCF classifications with their classes converted into the relevant PFTs.

0

2

4

6

8

10

12

NPP Sresp NBP

GLC2000

MODIS

VCF-MODIS

Fig 9: SDGVM carbon fluxes for GLC2000, MODISand MODIS VCF, (Gt y-1).

0

2

4

6

8

10

12

Roff EVT

GLC2000

MODIS

VCF-MODIS

Results are presented for the northern latitudes (lat ≥ 50)for the last ten years of CRU3 climatology: 1997-2006.

Fig 11: SDGVM Runoff and EVT for GLC2000, MODISand MODIS VCF, (Gt y-1).

•EVT was lower and subsequently Runoff increased for classifications with more herbaceous cover but not significantly.•Significant deviations in NBP were observed at high latitudes due to classifications differences in bare ground/herbaceous coverage.

Fig. 12: SDGVM NBP difference: MODIS VCF – GLC2000, (g m-2 y-1).Differences are located in latitudes>70°

Fig.11: SDGVM Evapotranspiration difference: MODIS VCF - GLC2000, (mm). Deviations are mainly located in agricultural regions were the model was driven withdifferent crop fractional cover.

•The land covers are significantlyDifferent but no clear reason to prefera specific one.•The different approach of MODIS-VCF means that it can beused to modify other land use products.

1) MODIS Product MCD45A1Source: MODIS sensor on-board the Terra and Aqua satellites.Type: Active Fire productSpatial Resolution: Global Coverage at 500 mTemporal Resolution: Monthly, 2000-2010Available Format: GeoTIFFObtained From: http://modis-fire.umd.edu/Burned_Area_Products.html

2) MODIS Product MOD14CMH & MYD14CMHSource: MODIS sensor on-board the Terra and Aqua satellites.Type: Fire Radiative PowerSpatial Resolution: Global Coverage at 0.5 degrees.Temporal Resolution: Monthly, 2002-2010Available Format: HDFObtained From: http://modis-fire.umd.edu/Active_Fire_Products.html

3) Global Fire Emissions Data (GFED) Emissions ProductSource: Derived mainly from MODIS images, land cover data sets and biochemical models.Type: Carbon EmissionsSpatial Resolution: Global at 0.5 degrees.Temporal Resolution: Monthly, 1997-2009Available Format: txtObtained From: http://www.falw.vu/~gwerf/GFED/

5) Global Fire Emissions Data (GFED) Burned Area ProductSource: Derived mainly from MODIS images.Type: Burned areaSpatial Resolution: Global at 0.5 degrees.Temporal Resolution: Monthly, 1996-2009Available Format: txtObtained From: http://www.globalfiredata.org/

4) Along Track Scanning Radiometer, World Fire Atlas (ATSR WFA)Source: ATSR/ATSR-2/AATSR instrument on-board the ERS-1/ERS-2/ENVISAT satellite.Type: Active Fires Product (night only)Spatial Resolution: Global at 1km.Temporal Resolution:Monthly,1996-2011Available Format:txtObtained From: http://wfaa-dat.esrin.esa.int/

Fig.13: Burned Area and Carbon Emissions according to GFED.Note that even though more carbon emissions originate fromEurasia, emissions per area burned is higher for N. America.

Fig 14: Comparison of burned area for N America between MODIS and GFED. Carbon emissions are shown on the secondary y-axis.Differences between the Burned Area products are attributed to thedifferent methodologies used.

Fig.15: Fraction of area burned for fire and non-fire years in N. America and Eurasia

•Significant temporal and spatial variations of fire activity.•Fire Return Interval in these latitudes is in the region of hundreds of years,an order of magnitude greater than the temporal range of the data sets.

•Fraction of area burned in models is strictly based on climatology (precipitation, fuel humidity).•Due to previous, models and especially CLM4.0 show significantly smaller variance compared to the data sets.•Models show good agreement of their long term mean of area burned compared to the data sets.•LPJ outputs considerable more carbon emissions, probably due to higher carbon content of burned vegetation.

Fig. 16: Burned Area and Carbon Emissions for different data sets and models

•Models cannot capture the stochastic nature of fires.•All 3 of our models burn the majority of the grid cells by a very small fraction, usually less than 1%.•Cumulatively they are in agreement with the data sets where few grid cells are burned by a considerably larger fraction.

Fig. 17: Fraction (%) of Area Burned for 3 models

•Fire data sets record is not long enough to drive the fire disturbances in models.•Instead we kept the fire disturbance algorithm of the SDGVM model but added the variance of the GFED Burned Area data set (SDGVM*).

Fig. 18: Burned Area (in mHc) for GFED,SDGVM and SDGVM* with the added variance.

Fig. 19: Carbon Emissions (Mt C) for SDGVM and SDGVM*

Fig. 20: NBP (Mt C) for SDGVM and SDGVM*

•The added variance of the burned area was subsequently transferred to the carbon emissions.•NBP didn’t exhibit any significant differences probably because even though variance increased, the % of area burned for each grid cell remained at levels lower than 3%.

Fig. 21: Biomass (kg m-2) for SDGVM and SDGVM*

Fig 22: BIOMASS Average 1997-2006: (Adjusted – Control) /Control x 100.

•Biomass is reduced with the adjusted higher variance fire regime.

•An entire new approach is required tomodel the fire regime, describing itsstochastic nature.•At this point models simply try tocapture the spatial average carbonemissions.•Describe the behaviour of the systemfor higher fractions of area burnedper grid cell.•GFED seems the most completeproduct since it offers both burnedarea and carbon emissions.