Mediterranean Ecosystems Worldwide: Climate Adaptation - Recent Advances and Issues in Environmental Science - page 32

conditions; (1) The winter must be wetter than the summers (>65% of the precipita-
tion falls in the winter half of the year), (2) the annual precipitation must not be too
low (>275 millimeters (mm)), (3) nor too high (<900 mm), (4) the winter must be cool
(<15°Celsius (C) mean temperature for the coldest month of the year), but (5) it can-
not have too much frost (<3% of the annual hours are below freezing). We used the
WorldClim [18] high resolution (2.5 arc-min or ~5 kilometer (km) horizontal resolu-
tion at the equator) grids of global climate data summaries from 1960 to 1990 to map
the current MCE where all fi ve Aschmann conditions are met.
Data for projections of future climate conditions were derived from the results
of the AOGCMs run to support IPCC's Fourth Assessment Report. The data [World
Climate Research Programme's (WCRP) Coupled Model Intercomparison Project
(CMIP) phase 3 multi-model dataset] include seven future emissions scenarios. Three
of these scenarios are used most often by modeling groups and are considered repre-
sentative of low (B1 or stabilization at 550 ppm atmospheric CO 2 ), moderate (A1b or
stabilization at 720 ppm atmospheric CO 2 ) and high (A2 or no stabilization) emission
trajectories [19]. We compiled the AOGCM output data for monthly surface air tem-
perature and precipitation fl ux for the 20th century and the 21st century for three future
emissions scenarios. While some modeling groups have generated multiple simula-
tions for a given scenario and others have done no simulations for a given scenario,
we analyzed all available AOGCM simulations in the CMIP multi-model dataset, in-
cluding 48 low emission simulations, 52 moderate emission simulations, and 36 high
emission simulations, for a total of 136 simulations of future climate. By doing so, we
treat each AOGCM simulation for a given emissions scenario as a unique and probable
experimental outcome and average the results, thereby elucidating a more robust set of
potential climate outcomes.
To reduce the variability associated with annual climate projections, we averaged
the monthly data in the AOGCM simulation results to two 30-year periods; one “cur-
rent” and one “future.” The WorldClim data is primarily derived from 1960 to 1990
weather records, so we averaged the monthly data from January, 1960 to December,
1989 for each of the modeled 20th century simulations to generate the current time pe-
riod. The majority of the model simulations end in 2100, so we averaged the monthly
data from January, 2070 to December, 2099 for each of the modeled 21st century simu-
lations to generate the future time period. We then subtracted the modeled current data
from the modeled future data to reduce modeling biases and generate projected cli-
mate anomalies. For example, an AOGCM simulation may have modeled the average
July temperature for a specifi c area to be 24°C for the current time period (1960-1989)
and 27°C for the future time period (2070-2099), so the projected climate anomaly
for that area would be 3°C. We used the change factor approach to downscale the
projected climate anomalies from the coarse resolution of the AOGCMs to the fi ner
resolution of the WorldClim data. This method involves interpolating the projected
climate anomalies and adding the interpolated data to the current climatology.
We applied Aschmann's [17] conditions to generate binary maps of the projected
future MCE for each AOGCM simulation. The size of the projected future MCE in
each region was compared to the size of the current MCE for each AOGCM simulation,