Consistent with its high elevation (>4000 m) and subtropical location (15°–25°S), the central Andes are expected to become warmer during the twenty-first century, affecting the population, ecosystems, and glaciers on the so-called South American Altiplano. Future changes in regional precipitation (even its sign) have been more difficult to estimate, partly because of the low resolution of current global climate models (GCMs) relative to the cross-mountain scale of the Andes. Nevertheless, summer season rainfall over the Altiplano exhibits a strong dependence on the magnitude of the zonal flow in the free troposphere, as quantified in this work using station data. Since GCMs indicate a consistent increase in westerly flow over the central Andes, hindering moisture transport from the interior of the continent, a simple regression analysis suggests a significant reduction (10%–30%) in Altiplano precipitation by the end of this century under moderate-to-strong greenhouse gas emission scenarios.
The South American Altiplano is a closed, high-level plateau (~4000 m MSL) located in the central Andes between 15° and 22°S. Bordered to the west by the Peru–Chile coastal desert and to the east by the humid Bolivian lowlands, the Altiplano exhibits a semiarid climate (Aceituno 1993; Garreaud et al. 2003), with annual mean precipitation ranging from about 600 mm in the northeast to <200 mm in the southwest. Most of the precipitation (>70%) occurs during austral summer [December–February (DJF); Fig. 1a] when mid- and upper-level easterly wind brings moist air that feed convective storms over the plateau (Garreaud 1999; Falvey and Garreaud 2005). Summer precipitation exhibits significant synoptic variability, largely explained by the local boundary layer moisture and zonal wind aloft (Garreaud 2001). The rest of the year the Altiplano is influenced by midlevel westerly flow that brings very dry air from the Pacific and precipitation events are nearly nonexistent (Vuille and Ammann 1997). At interannual time scales, there is a tendency for more (less) basinwide precipitation during La Niña (El Niño) years that is also largely explained by the intensity of the zonal flow aloft (Garreaud and Aceituno 2001; Vuille and Keimig 2004).
Given its semiarid climate and strong seasonality, future changes in summer precipitation and temperature over the Altiplano will affect water availability for human consumption, agriculture, glaciers, and ecosystems (e.g., Bradley et al. 2006). Global climate model (GCM)–based projections of climate change during the twenty-first century under the A2 scenario consistently show a free-tropospheric warming as high as +5°C over the central Andes (Bradley et al. 2006), and a simulation using a regional climate model (RCM) indicates a similar increase in surface air temperature (Urrutia and Vuille 2009). In contrast, expected changes in central Andes precipitation remain poorly determined in the Coupled Model Intercomparison Project phase 3 (CMIP3) GCMs runs (Christensen et al. 2007). Seth et al. (2010) and Thibeault et al. (2010) examined nine GCMs over the central Andes and found a tendency for a slight increase in precipitation during the rainy season and a slight decrease during the early season. In contrast, the RCM results by Urrutia and Vuille (2009) imply a decrease in summer precipitation over the central Andes south of 12°S.
To further illustrate the uncertainty in projections of precipitation, the vertical axis in Fig. 2 shows the rainfall change (ΔP) between the end of the century (2070–99) under the A2 scenario and the baseline period (1970–99) for 11 GCMs used in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) (see details in Table 1). Considering the Altiplano1 as the grid boxes with elevation >3000 m between 10° and 25°S, we calculated the area average and range (maximum to minimum) of the ΔP for each model. There is significant dispersion in the magnitude and even the sign of ΔP. Five models indicate an increase in precipitation and five indicate a decrease, resulting in an insignificant multimodel mean change of 1 mm month−1.
The lack of a robust trend in regional precipitation severely limits our capacity to foresee environmental changes over the Altiplano. To advance this issue, here we exploit the observed relationship between zonal wind aloft—a large-scale variable more reliable in GCMs—and Altiplano precipitation. Granted, this approach only allows us to project precipitation changes that are congruent with changes in zonal flow; however, given the strong coupling between these two fields over much of the Altiplano, we argue that such a signal is a significant fraction of the total change.
2. Zonal wind–precipitation relationship
The relationship between zonal wind aloft and precipitation over the Altiplano (easterly–wet, westerly–dry) holds at synoptic, seasonal, and interannual time scales. This relationship has mostly been derived using proxy data, such as cold cloud fraction (Vuille and Keimig 2004), outgoing longwave radiation (Garreaud and Aceituno 2001), glacier mass balance (Francou et al. 2003; Vuille et al. 2008), or very limited rainfall records (Falvey and Garreaud 2005). To our knowledge, however, only one study of Vuille (1999) has used actual precipitation observations from 6 meteorological stations over the plateau. Here we use monthly rainfall records from 108 stations in the central Andes above 3500 m (Fig. 2) from 1948 to 2007 (variable record length but at least 10 yr worth of data) from the Global Historical Climatology Network (GHNC version 2; Vose et al. 1992). Large-scale circulation was characterized using the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis from 1948 onward (Kalnay et al. 1996).
Figure 1b shows the correlation between monthly (DJF) zonal wind at the 200-hPa level (U200) in a grid box centered over the Altiplano (17.5°S, 70°W) and concurrent rainfall at each station. The result appears robust because all 108 stations exhibit negative P–U200 correlation, 88% of them significant at the 95% confidence level, including stations located within the plateau and farther north over the tropical Andes. The P–U200 correlations are largest (r2 ~ 0.6) along the western (dry) side of the Altiplano and decrease toward the eastern (humid) side. Because of the spatial coherence of the zonal wind field over the central Andes, very similar results are obtained if using U200 interpolated to each station’s coordinates. We also verified that monthly-mean values of U200 are strongly and negatively correlated with the number of days with easterly wind (u < 0) during the month and are not much influenced by few days with strong westerly or easterly flow, which explains the much better linear fit between zonal flow aloft and precipitation at monthly (or longer) time scales relative to results using daily values (Garreaud and Seluchi 2003).
3. Projected changes in upper-level winds
Projected changes in zonal wind are now examined using 11 CMIP3 GCMs (Table 1). Unlike precipitation, large-scale circulation is considered a reliable output from current GCMs (Christensen et al. 2007). Here we focus on the wind changes between the last 30 years of the A2 (2070–99) and Twentieth-Century Climate in Coupled Model (20C3M) (1970–99) scenarios. Figure 3a shows the multimodel summer mean difference in the 200-hPa wind field (ΔUV200) over the South American sector and adjacent oceans. Upper-level westerly flow becomes stronger (ΔU200 > 0) in two zonally elongated bands: one at subtropical latitudes (10°–30°S) and the other at high latitudes (poleward of 50°S). In between (~40°S) the westerlies slightly decrease in the future. Overall, no significant changes in the meridional flow or in the strength or position of the Bolivian high are apparent. The isobaric map is complemented by the latitude–pressure cross section of zonal wind differences at 70°W shown in Fig. 3b. Weakening (strengthening) of the tropospheric westerlies at midlatitudes (high latitudes) is consistent with the projected poleward shift of the SH storm track in a warmer world (Yin 2005). Near and above the tropopause, however, the westerlies increase at midlatitudes and a tongue of ΔU200 > 0 extends downward into the troposphere at subtropical latitudes, causing anomalous westerly winds over the central Andes at upper and midlevels. Easterly winds are expected to intensify only above the equatorial Andes.
The multimodel mean difference in zonal flow over the Altiplano is not too large (ΔU200 ~ +1–3 m s−1), but it is superimposed on a weak baseline easterly flow. The summer mean “divide” between midlevel easterly and westerly flow (i.e., the U = 0 m s−1 contour in Fig. 3b) is at 20°S in the current climate, but it could move as far north as 10°S by the end of the century in the A2 scenario. To gauge the consistency of the zonal wind differences over the Altiplano, let us look again at Fig. 2, whose horizontal axis shows, for each model, ΔU200 interpolated to 17.5°S, 70°W. In contrast to ΔP, all the models predict enhanced westerlies (or decreased easterlies) with a significant multimodel mean of ΔU200 = 2.5 m s−1. The weakening of the easterly flow over the Altiplano during the twenty-first century is also found in more benign emission scenarios, with a multimodel mean ΔU200 of 1.6 and 2.3 m s−1 for B1 and A1B, respectively. We also verified (in 7 of the 11 GCMs with daily data) that the enhanced summer mean westerlies in the future are associated with a decrease in the number of days with easterly wind.
The summertime enhanced westerlies at subtropical latitudes and the retraction of the easterlies into lower latitudes set the stage for drier than current conditions across much of the central Andes. What is the prospect for a shift in precipitation seasonality? To answer this question, we examined the multimodel mean annual cycle of U200 over the Altiplano calculated for the current and future climate (not shown). The westerly component becomes stronger year-round (but with a weak easterly difference in May–June) with the largest differences in summer. Indeed, the narrow window of time when U200 < 0 in the current climate (DJF) is absent in the future simulation that exhibits monthly-mean westerly winds over the Altiplano year-round suggests a shortening and weakening of the rainy season instead of a shift.
4. Projected changes in precipitation
We now estimate at each station i the wind-driven change in summer precipitation as , where ΔU200 is the multimodel mean interpolated at 17.5°S, 70°W and βi = r(Pi, U200)σ(Pi)/σ(U200) is the slope of the linear fit calculated using the observations in section 2. The results are shown in terms of absolute change (; Fig. 1c) and the difference relative to the current climatology (; Fig. 1d). Other alternatives to calculate (e.g., using ΔU200 interpolated to the station coordinates) produce very similar results.
Because of the weaker easterly flow, the central Andes exhibit a decrease in precipitation toward the end of the century that, averaged over all stations, equals 16.5, 15.3, and 10.8 mm month−1 for the A2, A1B, and B1 scenarios, respectively. The absolute change in A2 is as large as −30 mm month−1 along the western and central parts of the Altiplano (Fig. 1c) but decreases (<−10 mm month−1) on the eastern side, where the coupling between U200 and P is weaker. When considering the relative changes, a north–south gradient is also evident with differences from −10% in the northern sector down to −35% in the south (Fig. 1d). The decrease in precipitation over the Altiplano obtained from our multi-GCM regression is generally consistent (in sign and magnitude) with the projections by Urrutia and Vuille (2009) based on their single-RCM dynamical downscaling, but the projections do not support the tendency for a more intense rainy season found in the GCM-based analysis by Seth et al. (2010) and Thibeault et al. (2010). We have refrained from calculating precipitation changes in other seasons, because the very low precipitation leads to an insignificant P–U200 relationship. Nevertheless, as discussed before, westerly wind anomalies will prevail almost year-round in the future, increasing the westerly flow from fall to spring, thus hindering any expansion of the rainy season.
Finally, we verified that future climate scenarios satisfy the underlying hypothesis in the P–U200 relationship, namely, a marked zonal gradient in midtropospheric moisture across the central Andes. To this end, Fig. 4 shows the summer mean of the water vapor mixing ratio at 600 hPa at 17.5°S over the Pacific Ocean (75°W) and the Bolivian lowlands (65°W) for the current climate and the A2 scenario. In the current climate, almost all models capture the marked moisture gradient between the very dry Pacific sector (~2 g kg−1) and the more humid conditions to the east of the Andes. In the future climate, there is a generalized moistening at low latitudes (e.g., Held and Soden 2006) and the multimodel mean increases at both sides of the central Andes. Relative to the (simulated) current conditions, the multimodel mean increase is ~1.5 g kg−1 over the Bolivian lowlands but only ~0.5 g kg−1 over the Pacific sector. More importantly, to the west of the Andes remains below 3 g kg−1 in most of the GCMs, a value below the threshold for moist convection over the Altiplano (Falvey and Garreaud 2005), so the predicted westerly flow will continue hindering precipitation. Although in the current climate midtropospheric moisture over the lowlands east of the Andes has a slightly significant relationship with Altiplano rainfall in most stations, its effect is conditioned to the occurrence of easterly winds. Thus, the increase in moisture over the continent will hardly attenuate the wind-driven drying trend over the Altiplano.
5. Concluding remarks
The Altiplano region is poorly resolved in current GCMs given their relative coarse resolution relative to the narrow Andes. Not surprisingly, direct analysis of GCM outputs results in inconsistent, nonsignificant precipitation changes over the Altiplano, in contrast with a robust warming trend expected for the rest of the twenty-first century. To circumvent this problem, in this work we exploited a rather strong relationship between midtropospheric zonal winds and precipitation over the central Andes to project changes in regional rainfall for the end of the twenty-first century. In the current climate, Altiplano rainfall is largely concentrated in austral summer and an easterly–wet (westerly–dry) pattern has been found in synoptic, seasonal, and interannual time scales. The relationship was quantified by using more than 100 rain gauges along the central Andes and NCEP reanalysis winds. The linear correlations are largest (r2 ~ 0.6) along the western (dry) side of the Altiplano and decrease eastward toward the more humid side of the plateau. When examining the projected changes in free-tropospheric wind (a reliable output from GCMs), we found an almost year-round increase in westerly flow at mid- and upper levels over the central Andes, which results in a decrease of the moisture transport toward the Altiplano from the interior of the continent during summer, reducing the summer precipitation between 10% and 30% relative to current values. This approach only permits an estimation of the part of the precipitation change due to wind aloft change. However, the GCMs’ consistency of the ΔU200 sign and the strong relationship observed with wind and precipitation in the current climate indicate that this wind-driven summer precipitation change is a significant fraction of the total change. The rest of the year the Altiplano would continue as dry as in present conditions, which when added to the increase in surface air temperature (>3°C) signals a complicated prospect for the water resources in this semiarid region.
This investigation was supported by the FONDECYT Grant 3110120. NCEP reanalysis data were provided by the NOAA/OAR/ESRL PSD in Boulder, Colorado, from its Web site (www.esrl.noaa.gov/psd/). We acknowledge the modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) and the WCRP’s Working Group on Coupled Modelling (WGCM), for their roles in making available the WRCP CMIP3 multimodel dataset. Support of this dataset is provided by the Office of Science, U.S. Department of Energy.
Because of their coarse resolution, some of the CMIP3 GCMs do not have terrain elevation above 3000 m MSL in the central Andes and were not considered in our analysis.