Auxiliary Material Submission for the Paper

Mediterranean water cycle changes: transition to drier 21st

century conditions in observations and CMIP3 simulations

Annarita Mariotti

Earth System Science Interdisciplinary Center,

Univ. of Maryland, College Park and ENEA/Ente Nazionale per le Nuove Tecnologie,

l’Energia e l’Ambiente, Rome, Italy.

Ning Zeng

Dept. of Atmospheric and Oceanic Science and Earth System Science Interdisciplinary

Center, University of Maryland, College Park.

Jin-Ho Yoon

Dept. of Atmospheric and Oceanic Science,University of Maryland, College Park.

Vincenzo Artale

ENEA/Ente Nazionale per le Nuove Tecnologie, l’Energia e l’Ambiente, Rome, Italy.

Antonio Navarra

CMCC/Centro-Euro-Mediterraneo per i Cambiamenti Climatici and Istituto Nazionale di

Geofisica e Vulcanologia, Bologna, Italy.

Pinhas Alpert

Tel-Aviv University.

Laurent Z.X. Li

Laboratoire de Météorologie Dynamique, Paris, France.

Introduction

The material presented here consists of seven auxiliary figures (Fig.1S-7S) and two tables

(Table 1S-2S). The data used for their derivation is described in the section “Data and

methodology" of the paper. As additional information to this section we specify that the

time-series derived merging CRU and PRECL data (CRU/PRECL) over the period 1901-

2007 is the original CRU data for the period 1901-1947; for 1948-2007, it is constituted

of PRECL anomalies relative to the 1948-2002 climatology added to the CRU 1948-2002

climatology.

Figure 1S shows 20th century anomalies relative to 1950-2000 for: a) Ts averaged over

the Mediterranean land region as simulated by the CMIP3 models and in CRU data; b)

SST anomalies in the Mediterranean region from the CMIP3 models’ and HadISST data.

Data are six-years running means of annual mean anomalies. The comparison shows an

overall reasonable agreement between models and observations with progressively

increasing Ts/SST over the course of the 20th century. Also, as in the observations, the

simulated Ts increase is greater than the SST increase. However differences exist, most

notably the observed 20th century Ts/SST trends are larger in the observations than in

their simulated counterpart (specifically, CRU/HadISST show larger positive trends early

in the 20th century and since the mid-1970s). Other characteristics of observed Ts/SST

decadal variability are also not reproduced in the CMIP3 simulations.

Figure 2S displays Mediterranean land water cycle changes during the 1900-2100 period

based on the NCAR CCSM3 CMIP3 model runs (7 member ensemble). In addition to

precipitation and evaporation anomalies, runoff and soil moisture anomalies are also

displayed. A progressive long-term decrease in soil moisture and runoff are simulated

over the 1900-2100 period. Changes in runoff have important implications for land water

resources and for the salinity of the Mediterranean Sea [Rohling and Hilgen, 1991].

Figure 3S shows Mediterranean-mean precipitation, evaporation and precipitation minus

evaporation (P-E) annual anomalies for the period 2070-2099 relative to 1950-2000, for

individual model ensembles. Consensus is high among the 14 models which were

analysed, for both land and sea: all show a decrease in precipitation and in P-E.

Figure 4S displays regional precipitation, evaporation and precipitation minus

evaporation anomalies in 2070-2099 compared to 1950-2000 for the “wet” and “dry”

seasons. Here anomalies are expressed in units of standard deviation considering the

variance of individual model mean anomalies around the ensemble mean (a measure of

the uncertainty associated to the projection). Results show that anomalies are well-above

inter-model variance over the Mediterranean Sea.

In Figure 5S, regional annual mean precipitation, evaporation and precipitation minus

evaporation anomalies are displayed for the target periods 2020-2049 and 2070-2099.

Major features are common to the two periods with anomalies deepening in 2070-2099

compared to 2020-2049.

Figure 6S shows winter mean (DJFM) observed precipitation anomalies in the

Mediterranean region and the North Atlantic Oscillation (NAO) index [Jones et al., 1997]

over the period 1900-2007. The precipitation decrease during the period 1960-1990 has

been linked to the increase in the NAO index during the same period. It should be noted

that, while the NAO is the single greatest influence on Mediterranean winter

precipitation, especially on decadal time scales, other modes of variability also influence

regional precipitation. This is particularly true outside of the winter season.

Figure 7S displays CMIP3 regional projections of mean sea level pressure anomalies in

2070-2099 compared to 1950-2000 for the “wet” and “dry” seasons. During the wet

season projections show an increase in SLP over the whole Mediterranean region. During

the dry season SLP changes are mixed, with an increase in SLP in northern parts and a

decrease in southern parts.

Table 1S presents a validation of CMIP3-simulated 20th century precipitation, Ts and

SST over the Mediterranean region. The table reports CMIP3-simulated area-averaged

“annual”, “wet” season and “dry” season mean values during the period 1950-2000 for

SST, Ts and precipitation (the latter two are land-only averages) and associated biases

compared to observations (observational Ts and P are from CRU; SSTs are from

HadISST). Area-averages are for the whole Mediterranean region as defined in the paper

and for four sub-regions coinciding with the four quadrants: North West Med. (10ºW-

15ºE and 38ºN-47ºN), North East Med. (16ºE-40ºE and 38ºN-47ºN), South East Med.

(16ºE-40ºE and 28ºN-37ºN) and South West Med. (10ºW-15ºE and 28ºN-37ºN). Overall,

Mediterranean-mean biases are small for all time periods (about -3% in precipitation,

about -1K for Ts and about -2K for SST), with the CMIP3 ensemble mean being drier

and cooler than observed. There exists a Ts cold bias for all sub-regions and time-periods

with values up to 1.7K; the SST cold bias is generally 1-2K, but peaks in NE

Mediterranean region (including the Adriatic and Black Seas). Precipitation biases vary

for the various sub-regions, and are generally larger in Southern parts which tend to be

15%-30% wetter (drier) than observed during the “dry” (wet) season. In Northern parts,

biases are larger during the dry season (about -10%). These results, in agreement with

previous work by [Giorgi and Lionello, 2007].

Table 2S presents Mediterranean-averaged precipitation and P-E trends from 20th

century observations and CMIP3 model simulations for the periods 1900-2007, 2000-

2050 and 2000-2099. Over the period 1900-2007, both observed and simulated

Mediterranean-averaged land-precipitation display significant negative trends. The

negative precipitation trend significantly increases during the period 2000-2099 to about

-0.02 mm/d decade. During the same period, a significant negative trend also

characterizes P-E over both land and sea. P-E negative trends significantly increase in the

21st century: rate is -0.01 mm/d per decade over land and 0.04 mm/d per decade over the

sea.

References

Giorgi, F., and P. Lionello (2007), Climate change projections for the Mediterranean region, Global and Planetary Change. Jones, P. D., et al. (1997), Extension to the North Atlantic Oscillation using early instrumental pressure observations from Gibraltar and south-west Iceland, Int. J. Climatol., 17(13), 1433-1450. Rohling, E. J., and F. J. Hilgen (1991), The eastern Mediterranean climate at times of sapropel formation: a review, Geologie en Mijnbouw, 70, 253-264.

a

b

Figure 1S: 20th century anomalies relative to 1950-2000 for: a) Ts averaged over Mediterranean land

region from the CMIP3 models’ ensemble and in CRU data; b) SST anomalies in the Mediterranean region

from the CMIP3 models’ and HadISST data. CMIP3 ensemble mean is in black, individual model means

are in grey and observational data are in red. Data are six-years running means of annual mean anomalies

in degrees K.

Figure 2S: Mediterranean land water cycle changes, as in Figure 1, except based exclusively on the

NCAR CCSM3 CMIP3 model runs (7 member ensemble). Land-only averaged anomalies for precipitation

(blue), evaporation (brown) and runoff (black) in the Mediterranean region (left axis) are displayed along

with those of soil moisture (green; right axis) .

a

b

a

Figure 3S: Mediterranean mean precipitation (blue cross), evaporation (brown circle) and precipitation

minus evaporation (black triangle) annual mean anomalies for the period 2070-2099 relative to 1950-2000

for individual model ensemble means. Area-averages over the Mediterranean region are computed

separately for land (panel a) and sea (panel b).

a b

c d

e f

Figure 4S: Same as Figure 3 except anomalies are expressed in units of standard deviation considering the

variance of individual model mean anomalies around the ensemble mean.

a b

c d

e f

Figure 5S: As Figure 3 except anomalies are annual means for the periods 2020-2049

and 2070-2099 (panels a, c, e and b, d, f respectively).

Figure 6S: Winter mean (DJFM) precipitation in the Mediterranean region and the NAO index (reversed

sign) over the period 1900-2007. Precipitation is the average of CRU/PRECL and GHCN. Six-years

running mean anomalies are displayed.

aa b

Figure 7S: As Figure 3 except anomalies are of sea level pressure (units are hPa).

(a)

(b)

(c)

Table 1S: CMIP3-simulated 20th century precipitation, Ts and SST over the Mediterranean region and

associated biases. Reported are CMIP3-simulated area-averaged “annual”, “wet” season and “dry” season

mean values during the period 1950-2000 and associated biases compared to observations. Area-averages

are for the whole Mediterranean region (TotMed) as defined in the paper and for four sub-regions

coinciding with the four quadrants (NWMed, NEMed, SEMed, SWMed). In each cell, mean values (left)

and biases (right) are reported. (a) Land-only precipitation (P; mean values in mm/d; biases against CRU

data in %) (b) Land-only Ts (degrees K; biases against CRU data) (c) SST (degrees K; biases against

HadISST data).

-0.0452 +/- 0.0010-0.0389 +/- 0.0024-0.0026 +/- 0.0009P-Emod(sea)

-0.0101 +/- 0.0004-0.0103 +/- 0.0012-0.0026 +/- 0.0004 P-Emod(land)

-0.0191 +/- 0.0007-0.0201 +/- 0.0019 -0.0072 +/- 0.0007 Pmod

__-0.0048 +/- 0.0028Pobs

2000-20992000-20501900-2007

Table 2S: Long-term trends in Mediterranean-averaged land-precipitation (P) and precipitation minus

evaporation (P-E) for the periods 1900-2007, 2000-2050 and 2000-2099. These are based on GHCN

precipitation (Pobs), an average of CMIP3 model-runs for land-precipitation (Pmod) and P-E (P-Emod)

averaged separately over land and sea. 95% confidence intervals are shown.