Observations and Simulated Mechanisms of Elevation-Dependent Warming over the Tropical Andes

Oscar Chimborazo aDepartment of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York
bDirección de Investigación y Desarrollo, Facultad de Ingeniería Civil y Mecánica, Universidad Técnica de Ambato, Ambato, Ecuador
cYachay Tech University, School of Physical Sciences and Nanotechnology, Urcuquí, Ecuador

Search for other papers by Oscar Chimborazo in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0002-9736-4518
,
Justin R. Minder aDepartment of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York

Search for other papers by Justin R. Minder in
Current site
Google Scholar
PubMed
Close
, and
Mathias Vuille aDepartment of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York

Search for other papers by Mathias Vuille in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Many mountain regions around the world are exposed to enhanced warming when compared to their surroundings, threatening key environmental services provided by mountains. Here we investigate this effect, known as elevation-dependent warming (EDW), in the Andes of Ecuador, using observations and simulations with the Weather Research and Forecasting (WRF) Model. EDW is discernible in observations of mean and maximum temperature in the Andes of Ecuador, but large uncertainties remain due to considerable data gaps in both space and time. WRF simulations of present-day (1986–2005) and future climate (RCP4.5 and RCP8.5 for 2041–60) reveal a very distinct EDW signal, with different rates of warming on the eastern and western slopes. This EDW effect is the combined result of multiple feedback mechanisms that operate on different spatial scales. Enhanced upper-tropospheric warming projects onto surface temperature on both sides of the Andes. In addition, changes in the zonal mean midtropospheric circulation lead to enhanced subsidence and warming over the western slopes at high elevation. The increased subsidence also induces drying, reduces cloudiness, and results in enhanced net surface radiation receipts, further contributing to stronger warming. Finally, the highest elevations are also affected by the snow-albedo feedback, due to significant reductions in snow cover by the middle of the twenty-first century. While these feedbacks are more pronounced in the high-emission scenario RCP8.5, our results indicate that high elevations in Ecuador will continue to warm at enhanced rates in the twenty-first century, regardless of emission scenario.

Significance Statement

Mountains are often projected to experience stronger warming than their surrounding lowlands going forward, a phenomenon known as elevation-dependent warming (EDW), which can threaten high-altitude ecosystems and lead to accelerated glacier retreat. We investigate the mechanisms associated with EDW in the Andes of Ecuador using both observations and model simulations for the present and the future. A combination of factors amplify warming at mountain tops, including a stronger warming high in the atmosphere, reduced cloudiness, and a reduction of snow and ice at high elevations. The latter two factors also favor enhanced absorption of sunlight, which promotes warming. The degree to which this warming is enhanced at high elevations in the future depends on the greenhouse gas emission pathway.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Mathias Vuille, mvuille@albany.edu

Abstract

Many mountain regions around the world are exposed to enhanced warming when compared to their surroundings, threatening key environmental services provided by mountains. Here we investigate this effect, known as elevation-dependent warming (EDW), in the Andes of Ecuador, using observations and simulations with the Weather Research and Forecasting (WRF) Model. EDW is discernible in observations of mean and maximum temperature in the Andes of Ecuador, but large uncertainties remain due to considerable data gaps in both space and time. WRF simulations of present-day (1986–2005) and future climate (RCP4.5 and RCP8.5 for 2041–60) reveal a very distinct EDW signal, with different rates of warming on the eastern and western slopes. This EDW effect is the combined result of multiple feedback mechanisms that operate on different spatial scales. Enhanced upper-tropospheric warming projects onto surface temperature on both sides of the Andes. In addition, changes in the zonal mean midtropospheric circulation lead to enhanced subsidence and warming over the western slopes at high elevation. The increased subsidence also induces drying, reduces cloudiness, and results in enhanced net surface radiation receipts, further contributing to stronger warming. Finally, the highest elevations are also affected by the snow-albedo feedback, due to significant reductions in snow cover by the middle of the twenty-first century. While these feedbacks are more pronounced in the high-emission scenario RCP8.5, our results indicate that high elevations in Ecuador will continue to warm at enhanced rates in the twenty-first century, regardless of emission scenario.

Significance Statement

Mountains are often projected to experience stronger warming than their surrounding lowlands going forward, a phenomenon known as elevation-dependent warming (EDW), which can threaten high-altitude ecosystems and lead to accelerated glacier retreat. We investigate the mechanisms associated with EDW in the Andes of Ecuador using both observations and model simulations for the present and the future. A combination of factors amplify warming at mountain tops, including a stronger warming high in the atmosphere, reduced cloudiness, and a reduction of snow and ice at high elevations. The latter two factors also favor enhanced absorption of sunlight, which promotes warming. The degree to which this warming is enhanced at high elevations in the future depends on the greenhouse gas emission pathway.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Mathias Vuille, mvuille@albany.edu

1. Introduction

Climate change is expected to significantly modify the environmental services that mountain regions provide for human and natural systems, most of all as critical sources of freshwater (Viviroli et al. 2011; Huss et al. 2017). This situation is further exacerbated by the fact that in many mountain regions the highest elevations appear to be warming faster than the lower slopes or the surrounding lowlands (Pepin et al. 2015).

In the tropical Andes, the warmer climate and the associated rise of the freezing level height has led to a rapid decline of the glacierized surface area (e.g., Francou et al. 2003; Vuille et al. 2008; Bradley et al. 2009; Schauwecker et al. 2014, 2017; Yarleque et al. 2018). The implications of these changes in the Andean cryosphere are substantial, since populations living downstream from glaciated mountain ranges depend on the glacial meltwater, especially during the dry season and prolonged droughts (Kaser et al. 2010). Warming at high elevations also affects the dynamics of glacier- and permafrost-related natural hazards in the Andes, which may become more prevalent as permafrost thaws and lakes, dammed by moraines or bedrock, continue to form in front of the retreating glaciers. Landslides, gradual destabilization of slopes, and glacial lake outburst floods therefore increasingly pose major hazards for communities living down-valley (Carey et al. 2012; Cook et al. 2016; Drenkhan et al. 2019). Strong warming at high altitudes also affects the health and integrity of the Andean biosphere and high-elevation ecosystems. Several recent studies have documented changes in the composition of glacier-fed river and lake systems in the Andes of Ecuador due to the changing glacier melt contribution (Jacobsen et al. 2012; Cauvy-Fraunié et al. 2015; Quenta et al. 2016). Similarly, warming at high elevation in the Andes has led to upslope migration of the Andean tree line (Feeley et al. 2011; Lutz et al. 2013), changes in species composition of ecosystems (Michelutti et al. 2015; Morueta-Holme et al. 2015; Dangles et al. 2017; Zimmer et al. 2018; Moret et al. 2019), and emergence of new invasive species (Seimon et al. 2007).

A thorough understanding of how climate change will affect high-elevation sites in the tropical Andes, including whether future warming will indeed be enhanced at highest elevations, is thus of critical importance. Yet our understanding of the mechanisms and feedbacks that lead to enhanced warming at higher elevations is still rudimentary. Giorgi et al. (1997), in an early modeling study, documented that the enhanced warming over the Alps in their regional model was related to a snow-albedo feedback. Beniston et al. (1997) and Vuille and Bradley (2000), based on observations in the Alps and tropical Andes, suggested that enhanced warming might occur at higher elevations, but did not attribute this elevation-dependent warming (EDW) signal to a specific forcing or feedback mechanism. Vuille et al. (2003), in a subsequent study, noted a clear difference in the elevation dependence of the warming rate between the eastern and the western slopes of the Andes in both observations and climate models and suggested that changes in cloud cover might have played a role, but they did not perform a detailed mechanistic analysis of such a cloud cover feedback. Longwave and shortwave radiation are both affected by cloud cover changes, and the net radiative effect of clouds depends on their vertical distribution throughout the atmospheric column. Diagnosing such cloud cover feedbacks is therefore not straightforward and in the tropical Andes it is further complicated by the lack of long-term observations at high elevations that could provide a better understanding of the effects of clouds on EDW.

Rangwala and Miller (2012) and later Pepin et al. (2015), in two comprehensive reviews, discussed several additional hypotheses that could explain the EDW on a global scale, and suggested specific mechanisms and feedbacks that could be responsible for the observed increased warming at higher elevations in mountain regions. The enhanced free tropospheric warming in the tropics likely results from enhanced release of latent heat during condensation, thereby enhancing the warming of the middle and upper troposphere in the tropics (Bony et al. 2006; Ohmura 2012). Bradley et al. (2006), Urrutia and Vuille (2009), and Russell et al. (2017) all highlighted such enhanced warming at higher altitudes in the free troposphere along the Andean Cordillera in future projections. Furthermore, increased downward longwave radiation (DLR) due to the increase in the water vapor content of the upper troposphere, where initially the specific humidity (q) is lower and water vapor therefore has a much higher efficacy at absorbing and re-emitting longwave radiation may contribute to high-elevation warming (Rangwala et al. 2010, 2013). The radiative effect of absorption by water vapor is roughly proportional to the logarithm of its concentration and therefore a proportionally greater increase in DLR will take place at the highest elevations, where the atmosphere is colder and therefore the initial water vapor concentration is lower (Rangwala and Miller 2012; Chen et al. 2014; Pepin et al. 2015). It is important to note, however, that while changes in q affect climate, the initial increase in q is itself the result of a change in climate, mainly a warmer world; hence the final effect of q represents a positive feedback.

The snow-albedo feedback is also commonly invoked to explain enhanced warming at high elevations (e.g., Kotlarski et al. 2015; Walton et al. 2017; Winter et al. 2017; Minder et al. 2018). However, this feedback is generally not considered to be very relevant at high elevations in the tropics, since the tropical climate lacks thermal seasonality that would allow for the build-up of a “winter” snowpack. Instead snow and ice are restricted to the highest peaks, and snow falling outside glaciated areas usually melts within a matter of a few days (Wagnon et al. 2009). Nonetheless, warmer temperatures will lead to enhanced snowmelt and precipitation will increasingly fall as rain rather than snow near the snow line (Rabatel et al. 2013). Both these processes result in a decrease of the local albedo, thereby enhancing the absorption of solar radiation, which in turn will reinforce the initial warming.

Here we employ observations and a regional climate model (RCM) to test which of these feedbacks contribute to EDW today and under future emission scenarios over the Andes of Ecuador. The aims of this study are (i) to investigate whether an EDW signal can be identified in observations and an RCM over the Ecuadorian Andes, (ii) to analyze how EDW might affect the highest elevations by the middle of the twenty-first century in two different emission scenarios, and (iii) to examine which mechanisms might be responsible for the EDW. Specifically, we hypothesize that EDW over the Ecuadorian Andes can be explained, at least in part, by changes in surface net radiation due to changes in cloudiness.

The paper is organized as follows: section 2 presents the data and methods used to better understand the EDW effect over the Ecuadorian Andes, with a special emphasis on the methodology used to probe the role played by the cloud radiative effect (CRE). The results are presented in section 3, followed by a discussion in section 4. The paper ends with conclusions in section 5 that focus on future research avenues to reduce uncertainties and gain further insight into how different feedbacks affect EDW in the tropical Andes.

2. Data and methods

a. Observational data

We used daily data from the weather station network of the Instituto Nacional de Meteorología e Hidrología del Ecuador (INAMHI) for the period 1986–2005. Daily temperature (mean, maximum, and minimum) data were processed through a quality control filter to eliminate unrealistic outliers and systematic errors (Chimborazo and Vuille 2021). The final dataset consists of 21 stations with a median missing value percentage of 2.9%, 2.0%, and 2.0% for mean, maximum, and minimum temperature, respectively (Table 1). The daily time series were used to obtain annual means and then to calculate the long-term trends using both an ordinary least squares (OLS) regression and a robust linear regression model (RLM) method, which reduces the contribution of outliers (Seabold and Perktold 2010; RLM module). To identify if a trend is statistically significant, we tested the null hypothesis that the slope of the trend is not significantly different from zero using an F test. Trends were tested for significance at the 95% confidence level (p values < 0.05), unless indicated otherwise.

Table 1.

Location (elevation and coordinates) and percentages of missing values of daily mean, maximum, and minimum temperature time series from selected stations of the INAMHI network from 1986 to 2005. Only stations with less than 25% missing daily values were included in the study and are listed. Minimum, maximum, median, and interquartile range (IQR) for each of the three temperature variables are listed at the bottom.

Table 1.

In addition, we used the monthly high-resolution (0.5°) Climatic Research Unit (CRU TS v4.03; Harris et al. 2020) temperature data to validate the performance of the WRF model in representing temperature as a function of altitude, in particular by comparing the lapse rate between observations and the CFSR-WRF model output. However, it should be noted that CRU data have a very limited assimilation of Ecuadorian weather stations at high elevations (only three stations > 2000 m).

b. WRF model setup

We performed multiple simulations with the Advanced Research version of the Weather Research and Forecasting Model (AR-WRF), version 3.7.1 (Skamarock et al. 2008; Wang et al. 2016), with a horizontal grid spacing of ∼10 km × 10 km and 51 levels in the vertical reaching a top pressure of 10 hPa. The full model domain covers continental Ecuador (1.78°N–5.32°S, 84.35°–72.75°W; Fig. 1), but since our focus is exclusively on EDW, we restrict most of our analyses to the Andean domain above 500 m above sea level (ASL).

Fig. 1.
Fig. 1.

Model domains used in the different experiments. (a) The black square indicates the parent domain of 50-km horizontal resolution used in the GCM-forced simulations (CCSM4-WRF), centered at 1.77°S, 78.10°W. (b) The green square is the nested domain in the GCM-forced simulation, which is almost identical to the single domain in the Reanalysis-forced experiments (CFSR-WRF), also centered at 1.77°S, 78.10°W. The negligible difference between the nested and the single domain in CFSR is due to the downscaling step used in the nested domain, which results in small differences in the coordinates values (less than ∼10−5 degrees, ∼10−3 km). (c) The red square represents the study domain for CFSR-WRF and CCSM4-WRF experiments. Model topography shown is interpolated to 50- and 10-km resolution for the parent and nested domain, respectively. Red dots indicate location of stations used in this study and listed in Table 1.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

WRF has been widely used to study climate variability and change over mountain regions (e.g., Qian et al. 2010; Salathé et al. 2010; Rasmussen et al. 2011, 2014; Letcher and Minder 2015; Walton et al. 2017), including over the Andes (Ochoa et al. 2014, 2016; Junquas et al. 2018; Heredia et al. 2018), as well as for studying EDW mechanisms and feedbacks (Gao et al. 2018; Minder et al. 2018). To our knowledge, however, no prior studies analyzed EDW mechanisms using regional climate models over the Andes of Ecuador.

Here we present results from four different simulations that were each run for 30 years, although our analysis is based on 20-yr time periods, in order to be consistent with how results from phase 5 of the Coupled Model Intercomparison Project (CMIP5) were analyzed in AR5. Results are qualitatively similar if the full 30 model years are used. The first of these simulations, CFSR-WRF, was driven by the NCEP Climate Forecast System Reanalysis (CFSR; Saha et al. 2010a,b). The simulation covers the period 1979–2010, but our analysis focuses on the historical period 1986–2005, as defined by IPCC (2013). It serves to represent present-day climate over Ecuador in a more realistic way than simulations driven by GCMs (e.g., McGlone and Vuille 2012; Posada-Marín et al. 2019).

The other three simulations, henceforth referred to as CCSM-WRF, were driven by boundary conditions obtained from the Community Climate System Model version 4 (CCSM4). These boundary conditions are based on a postprocessed product known as the NCAR CESM Global Bias-Corrected CMIP5 Output to Support WRF/MPAS (Model for Prediction Across Scales) Research (Monaghan et al. 2014). This bias-corrected dataset corresponds to CCSM4 member 6 (Bruyère et al. 2015) and produces temperature and precipitation changes over Ecuador that are close to the CMIP5 multimodel mean and median, as discussed below. The first of the three CCSM4-WRF simulations was run for the period 1975–2005, but again we limit our analysis to the time period 1986–2005. This simulation serves as a present-day control run (CCSM4-WRF CTRL). The other simulations were run following the representative concentration pathways (RCPs) 4.5 (moderate emission scenario) and 8.5 (high emission scenario; van Vuuren et al. 2011). They cover the years 2040–70, but here we analyze the period 2041–60, so results are comparable with the CMIP5 results in the IPCC AR5 (2013). This midcentury time period was chosen as a compromise between studying a period toward the end of the twenty-first century when the anthropogenic climate change signal has clearly emerged above the natural background climate variability, and a more recent period that is more relevant from an adaptation perspective for water managers, decision-makers, and policy makers (e.g., Buytaert et al. 2010; Vuille 2013). To allow soil fields to spin up, the first year in each simulation was discarded, as suggested by Bruyère et al. (2015).

A Student’s t test was applied to test for statistical significance at the 95% level between present-day climate and future projections, to ensure that differences are not the result of random internal variability. The significance of the EDW signal was assessed by testing the null hypothesis that the mean surface lapse rate is the same in the current and future climate. We performed a linear regression analysis of the temperature change versus elevation, separately for eastern and western slopes. The analysis of covariance (ANCOVA; Snedecor and Cochran 1989) with an interaction term (ΔT × slope) was used to analyze whether the two slopes are statistically significantly different from zero, but also from one another in the two emission scenarios.

This significance testing does not address model configuration uncertainty, but we emphasize that the focus of this study is on understanding EDW-related mechanisms, rather than making specific projections of future changes in climate. In addition, model configuration uncertainty was considered by comparing the performance of CCSM4 ensemble member 6 with a large CMIP5 multimodel ensemble over Ecuador, as discussed below.

The CFSR-WRF simulation only required a single domain to achieve the downscaled target grid spacing of 10 km, since the CFSR boundary conditions already provide a fairly fine grid spacing of 50 km. For the CCSM-WRF simulations, however, a two-step downscaling process was needed, given the coarser grid spacing of approximately 1° (∼111 km) in CCSM. This was achieved by defining a parent domain of 50-km grid spacing and a nested domain of 10 km (Fig. 1).

Our WRF simulations are based on a set of parameterizations that have been implemented in prior WRF studies over the Andes (e.g., Mourre et al. 2016; Moya-Álvarez et al. 2018; Martínez-Castro et al. 2019; Saavedra et al. 2020). To assess which parameterizations result in the best model performance over the Andes of Ecuador, Chimborazo (2018) carried out a series of sensitivity tests with different parameterization combinations. Based on these results and guided by prior studies in the region, we implemented the NCAR Community Atmosphere Model (CAM) version 5.1 microphysics scheme (Eaton 2011; Neale et al. 2012), and the New Tiedtke cumulus scheme (Tiedtke 1989; Zhang et al. 2011). The longwave and shortwave radiation are solved by fast versions of the Rapid Radiative Transfer Model (RRTM; Mlawer et al. 1997; Wang et al. 2016). The implemented planetary boundary layer scheme is from Yonsei University (Hong et al. 2006; Hu et al. 2010). The implemented Unified Noah land surface model (Tewari et al. 2004; Chen and Dudhia 2001) was used in simulations analyzed by Minder et al. (2016, 2018) to study EDW and albedo feedbacks in mountain regions. The selected atmospheric surface layer scheme is based on the Monin–Obukhov similarity theory, which has also been used in WRF simulations over the tropical Andes in prior studies (Junquas et al. 2018; Campozano et al. 2020). A summary of the model setup with the implemented parameterizations is listed in Table 2.

Table 2.

Physical parameterizations implemented in the simulations of WRF version 3.7.1.

Table 2.

Since the result of any WRF simulation depends on the driving model, it would be desirable to analyze an ensemble of WRF simulations driven by several CMIP5 models (Taylor et al. 2012). Given the computational constraints, however, this was not feasible. Instead, we compared future projections of temperature and precipitation from 40 models over nine subregions of Ecuador under the RCP8.5 scenario for the period 2041–60 (Table 3). To avoid biasing the results toward models with multiple ensemble members, we employed ensemble member 1 whenever more than one run was available for a given model. Figure 2 shows that CCSM4 member 6 simulates modest future changes in temperature and precipitation, indicating that this may be a rather conservative choice. Nonetheless, the simulated changes are close to the median and mean values of the CMIP5 multimodel ensemble over Ecuador. Hence, we consider our results to be representative of a broader multimodel ensemble, as member 6 of CCSM4 is unlikely to produce results that would be considered outliers in a multimodel ensemble of WRF simulations.

Fig. 2.
Fig. 2.

Projected change in annual mean precipitation (%) vs annual mean temperature (°C) over Ecuador from 40 CMIP5 models (see Table 3), resampled to a common resolution of 2°. Projections are based on RCP8.5, averaged over the period 2041–60, compared to reference period 1986–2005. The nine quadrangles cover the area of Ecuador. Results from 40 model simulations are shown with black circles. Pink square and green diamond symbols indicate the median and mean of the 40 models, respectively. The orange triangle represents the CCSM4 ensemble member 6 used to drive the CCSM4-WRF simulation.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Table 3.

The 40 CMIP5 models used to produce Fig. 2.

Table 3.

Chimborazo and Vuille (2021) performed a detailed quantitative analysis and validation of the performance of the WRF runs used in this study by comparing the spatiotemporal variability of precipitation and temperature in WRF, when driven by both CFSR and CCSM, with several observational datasets (CRU TS v4.03, CHIRPS v2.0, and in situ station data from the INAMHI network). Their analysis showed that the WRF simulations used here accurately simulate temperature and precipitation variability on daily, seasonal, and interannual time scales and the WRF-simulated lapse rate is consistent with observations (Navarro-Serrano et al. 2020). In their study Chimborazo and Vuille (2021) also pointed out that the model shows a dry bias at lower elevations and overestimates precipitation along the eastern Andean slopes. Bias-corrected future projections indicate drying along coastal areas in RCP4.5 and increased future precipitation along the eastern Andean slopes in both RCP4.5 and RCP8.5 scenarios by the middle of the twenty-first century (Chimborazo and Vuille 2021).

c. Calculation of the cloud radiative effect

One of the main objectives of this study is to better understand the potential role of clouds in enhancing warming at high elevations. Here, this influence was diagnosed by defining the cloud radiative effect (CRE) following Boucher et al. (2013), Goosse (2015) and Hartmann (2016):
CRE=RFallskyRFclearsky,
where RF is the radiative flux, regardless of whether it be the shortwave (SW), longwave (LW), or net radiation flux at the surface. All-sky and clear-sky represent the instantaneous fluxes calculated by the model’s radiative transfer scheme taking into account or neglecting the presence of clouds, respectively. The CRE therefore, indicates the effect that clouds exert on the radiative fluxes.

d. Calculation of the cloud water path and net radiation at the surface

To investigate the effect of clouds on the EDW, we calculate the cloud water path (CWP) as the vertical integral of the cloud liquid water mixing ratio (qcloud) plus the ice water mixing ratio (qice) over each grid cell, from surface to the top of the model:
CWP=1gp0ptop(qcloud+qice)dp,
where g is the acceleration of gravity, p0 is the pressure at surface, and ptop is the pressure at the top of the model. While in the literature the discussion of water vapor feedbacks is usually framed around the role of specific humidity, here our focus is on characterizing the role of the cloud water path, where the use of the mixing ratio, rather than specific humidity, is justified.
We also analyze the model-projected changes in the mean state of the all-sky and clear-sky net radiation at the surface (QS). Hartmann (2016) defines QS as follows:
QS=SWSSWS+LWSLWS,
where SWS and SWS are the downward and upward shortwave radiation fluxes at the surface, likewise LWS and LWS are equivalent indices for the longwave fluxes.

3. Results

a. Observed EDW over the Ecuadorian Andes

Observed temperature trends between 1986 and 2005 based on the INAMHI station network are shown in Figs. 3a–c and listed in Table 4. The calculated trends are not significantly different between the two regression approaches (see section 2a); therefore, only the RLM trends are shown and discussed here. Binning the data into 1000-m elevation bands before calculating trends (Figs. 3d–f), allows for a general assessment regarding the existence of an elevation-dependent warming. Bands are located 500 m apart and enclose all stations in a range of ±500 m; hence there exists a 500-m overlap from one band to the next. No data were available in the 1000–2000-m elevation band.

Fig. 3.
Fig. 3.

Temperature trends derived from the INAMHI station network. Decadal trends (K decade−1) of (a) mean, (b) maximum, and (c) minimum temperature for selected stations spanning the period 1986–2005. Dotted contours show the height of the terrain, starting at 1.5 km, at intervals of 1 km. Upward (downward) pointing triangles indicate positive (negative) trends that are statistically significant at p < 0.05. Circles indicate nonsignificant trends. (d)–(f) Trends as a function of elevation for mean, maximum, and minimum temperature using the same station information as in the above panels. The vertical bars indicate the elevation range for which the calculated trend is valid and the horizontal bars represent the 95% confidence interval of the estimate of the trend (slope) calculated from the linear regression. The numbers in parentheses to the left of the bottom panels indicate the number of stations that fall into each elevation band. Note the lack of data in the elevation band between 1000 and 2000 m. The total number of stations (in parentheses) does not equal the actual number of stations used, due to overlap of the elevation bins.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Table 4.

Temperature trends (K decade−1) at INAMHI stations for the period 1986–2005. Trends were calculated by ordinary least squares (OLS) regression and a robust linear model (RLM), but only RLM results are shown here. Statistically significant trends (p < 0.05) at individual stations based on an F test are shown in bold. Mean and maximum temperatures display warming rates that are significantly (p < 0.05) increasing with elevation.

Table 4.

Figures 3a–c show mostly positive trends over the Andes, consistent with recent studies (Vuille et al. 2015; Morán-Tejeda et al. 2016). While a clear EDW effect cannot be identified based on these three figures alone, stations with significant warming in mean and maximum temperatures are restricted to the Andean region, except for one station in the Amazon (Figs. 3a,b). For minimum temperature, a few stations in the coastal area also show a significant increase (Fig. 3c). The analysis of temperature trends at different altitudes, with data binned into 1000-m elevation zones (Figs. 3d–f), reveals that both trend magnitudes and their significance (trends where horizontal bars do not intersect with 0 K decade−1 line are significant at 95% level) vary, but there is a general tendency toward enhanced warming at higher elevations. Indeed, regressing temperature trends versus elevation reveals that mean temperature trends increase significantly (p < 0.05) as a function of elevation (0.08 K decade−1 km−1), while this elevation dependence is not significant for maximum and minimum temperature trends. When this analysis is repeated with temperature trends observed at individual stations (listed in Table 4), rather than using binned data, both mean and maximum temperature trends show a significant (p < 0.05) increase with elevation (0.08 and 0.11 K decade−1 km−1, respectively), while the elevation dependence of temperature trends for minimum temperature remains insignificant. However, because of the limited number of stations available for the trend calculation, and the lack of data in the 1000–2000-m elevation range, these results are characterized by large uncertainties.

b. EDW over the Ecuadorian Andes simulated by WRF

1) Historical simulations

A validation of the WRF simulations used here is presented in Chimborazo and Vuille (2021). Figure 4 further shows that the lapse rate calculated using the gridded dataset from CRU is similar to the one in the CFSR-WRF and CCSM4-WRF CTRL simulations, although caution should be exercised when comparing the results with the lapse rate from CRU, as this dataset does not contain any data above 3500 m. These results are comparable with those obtained by Urrutia and Vuille (2009), although they investigated a different period (1961–90) and a larger region. Nonetheless the lapse rate they determined based on CRU v 2.0 is identical to the one calculated here (4.9 K km−1) with CRU v 4.03, and the lapse rate (5.4 K km−1) of the regional model they used (PRECIS) is similar to the one in CCSM4-WRF CTRL (5.5 K km−1). Furthermore, the lapse rate determined from in situ station data by Urrutia and Vuille (2009) is the same as in our CFSR-WRF simulation (5.2 K km−1). Navarro-Serrano et al. (2020) also reported similar minimum and maximum temperature lapse rates, although they highlighted that significant differences exist between wet and dry season and that ENSO can introduce interannual variability in the lapse rate.

Fig. 4.
Fig. 4.

Temperature vs elevation of the terrain in (a) CRU data, (b) CFSR-WRF simulation, and (c) CCSM4-WRF CTRL simulation. The time period for the data used in (a)–(c) is from 1986 to 2005. The brown squares represent temperature below 500 m MSL and were not included in the calculation of the lapse rate.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Figures 5a–c show the surface temperature trends in the Andes obtained from CFSR-WRF over the 1986–2005 time period. For mean and maximum temperature, the warming is stronger on the western slopes, with more moderate warming on the eastern side. In the case of maximum temperature, the warming on the lower eastern slopes does not reach statistical significance. For minimum temperature, this east–west difference is less pronounced, with stronger warming at higher elevations. The surface temperature trends obtained from CCSM4-WRF CTRL show a slightly stronger dependence of the trends on elevation (Figs. 5d–f), especially for maximum temperature (Fig. 5e). Overall, however, the warming is reduced in CCSM4-WRF CTRL when compared to CFSR-WRF, and significant warming is generally restricted to areas above 2500 m. The CCSM4 forcing stems from a coupled ocean–atmosphere model and therefore will exhibit a different phasing of internal multidecadal variability, such as the Pacific decadal oscillation, than what occurred in reality. The CFSR reanalysis product, on the other hand, is based on assimilated observations and therefore communicates the actual observed multidecadal variability to WRF, which could explain the different warming patterns seen in the two simulations.

Fig.  5.
Fig. 5.

Decadal trends of (a) mean, (b) maximum, and (c) minimum temperature from the CFSR-WRF simulation, and (d) mean, (e) maximum, and (f) minimum temperature trends from the present-day CCSM4-WRF simulation. All trends span the period 1986–2005. The red contour lines indicate the 500-m altitude. The black contour lines start at 1500 m, with intervals of 1000 m. The green lines indicate the altitude of 4000 m. Hatching indicates regions with significant temperature trends at the 95% level based on a Mann–Kendall test.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

2) Future projections

Next, we analyze the projected future changes in temperature between the periods 2041–60 and 1986–2005 as a function of elevation. The high spatial resolution of our simulations allows distinguishing between temperature trends along the eastern and western slopes of the Andes. Analyzing the EDW separately for each side of the Andes makes sense, since they are influenced by different large-scale forcings: the western slopes are strongly influenced by Pacific sea surface temperature, while the eastern side is more strongly affected by convective activity and moisture transport over the Amazon basin (Vuille et al. 2000).

Figures 6a and 6b show that the projected future increase in mean temperature in the Andes is strongest at higher elevations in both the RCP4.5 and RCP8.5 scenarios, and that this effect appears to be more pronounced on the western than the eastern side.

Fig.  6.
Fig. 6.

(a) Projected temperature change in CCSM4-WRF simulation for the RCP4.5 scenario (2041–60 minus 1986–2005). (b) As in (a), but for the RCP8.5 scenario. In (a) and (b), the red contour lines indicate the 500-m altitude. The black contour lines start at 1500 m, with intervals of 1000 m. The green lines indicate the altitude of 4000 m. The thick blue line divides the western and eastern sides of the Andes for the analyses that are performed separately for each slope. No hatching was applied to this figure as temperature differences are significant at the 95% level everywhere.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

This is confirmed by Fig. 7, which depicts the projected future warming as a function of elevation, separately for each slope. The projected warming is significant (p < 0.05) at every 500 m elevation bin on both sides of the Andes in both scenarios. The ANCOVA analysis indicates that warming rates increase significantly as a function of elevation (p < 0.01) on both slopes and in both scenarios, but also that they are significantly different from one another (p < 0.01) for both the RCP4.5 and RCP8.5 scenarios. Hence the increase of the warming with elevation is more pronounced on the western side (RCP4.5: 0.08 K km−1; RCP8.5: 0.12 K km−1) than on the eastern side (0.06 K km−1 in both scenarios). Indeed, the projected future warming is larger on the western side at elevations above 2500 m, but weaker below that elevation, suggesting that other factors such as the role of continentality or maritime effects become more dominant at low elevation, away from the main Andean ridge (Fig. 7). The warming on the western side of the Andes is consistent with results in Urrutia and Vuille (2009), who simulated a significant warming under a high-emission scenario (A2) by the end of the century, with a clear EDW effect. However, in the same study the EDW effect on the eastern side was restricted to elevations above 2000 m, as the western Amazon basin saw even stronger warming in their projections. Whether these differences are due to the different regional and global models implemented, the different spatial domains analyzed, the different emission scenarios used, or the different time periods considered is not clear.

Fig.  7.
Fig. 7.

Temperature projections vs altitude for western (orange dots) and eastern (green dots) slopes of the Andes using the (a) RCP4.5 and (b) RCP8.5 scenarios. The data are binned into 500-m elevation bands, represented by the vertical bars. The horizontal bars represent the 95% confidence interval of the trend. The number of grid cells used for each band are indicated by the orange numbers in (a) for the western side and by green numbers in (b) for the eastern side. The eastern dots are shifted 50 m vertically for clarity. Note the different temperature scale of the x axis in (a) and (b). Warming rates increase significantly as a function of elevation (p < 0.01) on both slopes and in both scenarios.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

c. Free-tropospheric warming

Figure 8 shows the projected future change in tropospheric temperature and the mean zonal and vertical atmospheric circulation under both emission scenarios in an east–west section across the Andes. Clearly the warming of the free troposphere projects onto the surface warming seen on both slopes, with stronger warming at higher elevations. Additionally, in both RCP4.5 and RCP8.5 scenarios, enhanced easterly winds in the midtroposphere (∼3–6 km) generate enhanced subsidence on the leeward side of the Andean ridge. This strengthening of the midtropospheric easterlies is consistent with projected changes in the CMIP5 ensemble, which also shows strengthening of the midtropospheric easterlies along the equator (Collins et al. 2013). The enhanced subsidence on the leeward side of the Andes contributes to enhanced adiabatic warming as air masses descend, thereby promoting a warming asymmetry between the two sides of the Andes.

Fig.  8.
Fig. 8.

Vertical cross sections at latitude 0.07°N of the projected future changes in temperature, zonal wind, and vertical motion from CCSM4-WRF, for the (a) RCP4.5 and (b) RCP8.5 scenarios. The selected latitude includes the grid cell with the highest elevation in the model domain. The z component of the changes in the wind vectors is scaled to units of decimeters per second (dm s−1). Temperature changes are statistically significant at the 95% level everywhere in both projections, hence no hatching was applied. Wind vectors are only plotted where either the zonal or vertical component is significantly different at the 95% level.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

This mechanism is confirmed by an analysis of potential temperature (Fig. 9), indicating that the enhanced subsidence leads to adiabatic warming, which in turn is enhanced by the large increase in potential temperature at higher altitudes in the free troposphere. Furthermore, the stronger midelevation (∼1–3 km) warming over the Amazon region compared to the Pacific side (Fig. 8) also contributes to the weaker EDW signal found on the eastern side of the Andes.

Fig.  9.
Fig. 9.

Vertical cross sections at latitude 0.07°N of changes in water vapor mixing ratio and potential temperature from CCSM4-WRF simulations. (a) The mean state of water vapor mixing ratio for present-day (CTRL) simulation. The black contour lines represent the mean state of the potential temperature. (b),(c) The projected future changes of the water vapor mixing ratio for RCP4.5 and RCP8.5 scenarios, respectively. The black contour lines represent the changes in potential temperature. Both changes in water vapor mixing ratio and potential temperature are statistically significant at the 95% level everywhere in both projections, hence no hatching was applied.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

The enhanced subsidence over the high western slopes should favor drying of the air masses surrounding the peak elevations, and hence result in a reduced increase in water vapor content and therefore reduced cloudiness. Indeed, Fig. 9 shows that there is a smaller increase in the water vapor mixing ratio in RCP4.5 and RCP8.5 simulations in the immediate lee of the Andean ridge. This anomaly appears to be the result of the change in atmospheric circulation with enhanced subsidence as expressed by the potential temperature anomaly in the lee (westward side) of the Andean ridge. How this change in the atmospheric circulation affects cloud cover and hence the net radiation at the surface is investigated next.

d. The role of clouds: Changes in net radiation at the surface

Figures 10a–d show the projected changes of all- and clear-sky QS for both scenarios. There is a clear topographic effect over the eastern and western cordilleras with the all-sky QS increasing more strongly in the future above ∼2000 m (Figs. 10a,b). There is also a clear contrast at lower elevations, with QS decreasing on the northeastern side but increasing on the western side, especially in RCP4.5. This scenario shows a larger future precipitation reduction at low elevation over the southwestern flank, resulting in reduced cloud cover and a larger ΔQS (Chimborazo and Vuille 2021). The projections in the clear-sky QS (Figs. 10c,d) do not show this topographic signal, indicating that it is indeed a result of the cloud radiative forcing. In general, the clear-sky signal is much more homogeneous, with little difference between low and high elevations, except for the decrease in the RCP4.5 scenario below 1500 m over the western slopes. Over the highest peaks a stronger positive QS change is apparent in the clear-sky case, related to a surface albedo feedback due to a reduction in snow cover (see section 3e).

Fig.  10.
Fig. 10.

Projections of future net radiation change at the surface. (a),(b) Projections for all-sky conditions under RCP4.5 and RCP8.5 scenarios compared to the CTRL scenario, respectively; (c),(d) the same projections for clear-sky conditions. The red contour lines represent the 500-m elevation, while the black solid contour lines indicate the elevation at intervals of 1 km, starting at 1500 m. The green solid contour lines indicate the 4000-m altitude. Hatching indicates regions where net radiation changes are significant at the 95% level based on a Student’s t test.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Next, we consider the influence of the CRE on QS. Figure 11a shows the net CRE on longwave and shortwave radiation for the CTRL simulation. The overall CRE is negative throughout the domain analyzed, indicating that the cloud-induced reduction in shortwave incoming radiation outweighs the increase due to enhanced longwave radiation. It is noticeable, however, that this negative effect is enhanced along the Andean slopes between ∼500 and ∼3500 m, indicating a stronger influence of the clouds in reducing surface net radiation at these elevations than at the highest altitudes. Figures 11b and 11c show the projected future changes of the CRE (ΔCRE) for RCP4.5 and RCP8.5 scenarios, respectively. In both scenarios, ΔCRE is generally positive for altitudes > 2500 m leading to an increase in QS in the future. However, ΔCRE is negative over some central Andean regions (between ∼2500 and ∼3500 m), which correspond to inter-Andean valleys. On the lower slopes of the Andes, ΔCRE is negative, suggesting that clouds will more strongly reduce surface net radiation in the future, except for the lowest elevations on the western slopes in RCP4.5. Figure 11d shows the elevation dependence of the CRE in the CTRL run. The eastern side of the Andes presents a weak tendency toward more-negative values of the CRE with increasing elevation, while the elevation dependence on the western side is affected by strong scatter and very negative values at low altitudes over northwestern Ecuador. Figures 11e and 11f indicate a nonlinear increase of the CRE with elevation for both scenarios and on both sides of the cordillera, especially between ∼1500 and ∼3500 m. The linear regressions in these figures give a general sense of the trend as a function of altitude, but the trend is clearly not linear and above ∼4000 m ΔCRE is again slightly reduced.

Fig.  11.
Fig. 11.

Cloud radiative effect (CRE) for net radiation at the surface (QS). (a) Difference between the all-sky minus the clear-sky net radiation at the surface in the present-day simulation (CTRL). (b),(c) Projections of the CRE for RCP4.5 and RCP8.5 scenarios, respectively. Hatching indicates regions where cloud radiative effect changes are significant at the 95% level based on a Student’s t test. The red contour lines represent the 500-m elevation, while the black solid contour lines indicate the elevation at intervals of 1 km, starting at 1500 m. The green lines indicate the altitude of 4000 m. The magenta line shows the division between the western and eastern sides of the Andes used in this study. (d) CRE as a function of elevation for western (orange dots) and eastern (green dots) sides of the Andes for the CTRL simulation presented in (a). (e),(f) As in (d), but for the projected changes in the RCP4.5 and RCP8.5 scenarios, respectively. The solid lines represent a linear regression for eastern (green) and western (orange) slopes. The black solid line represents the regression combining the data from both sides of the Andes.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Next, we investigate the relative contributions of the CRE on net shortwave and longwave radiation, respectively. Figure 12a shows the CRE on the net shortwave radiation at the surface (CRESWs) for the present-day climate (CTRL simulation), with negative values throughout the domain. This implies that cloud cover leads to a reduction of the absorbed solar radiation at the surface since the albedo is enhanced by the presence of clouds. Figures 12b and 12c show the projected future changes of CRESWs. The positive changes in most of the high mountains indicate a reduced influence of clouds on the net shortwave radiation flux at those elevations in both scenarios, although RCP8.5 shows a stronger signal at higher elevations. Figure 12d shows the elevation dependence of the CRE in the CTRL run. On the eastern side of the Andes CRESWs becomes more negative with increasing elevation. The linear regression for the western side is strongly affected by very negative values at lower elevations over northwestern Ecuador, producing a positive linear dependence on elevation. Figures 12e and 12f show the elevation dependence of the future changes of CRESWs for both scenarios. The eastern side shows a clear tendency for positive changes of CRESWs with elevation, while in RCP4.5, large scatter at low elevation affects the linear regression of the western side, resulting in a very weak elevation-dependence. In the RCP8.5 scenario, however, both sides of the Andes exhibit increasingly positive CRESWs changes with higher elevations. Overall, these results indicate that under future warming, a weakening of the negative CRESWs over the Andes contributes to the increase in high-elevation QS shown in Fig. 10.

Fig.  12.
Fig. 12.

As in Fig. 11, but for cloud radiative effect on net shortwave radiation at the surface (CRESWs).

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

The CRE on the net longwave radiation at the surface (CRELWs; Fig. 13a) is positive throughout the entire domain, consistent with the observed CRE on the longwave radiation at the top of the atmosphere (Boucher et al. 2013; Hartmann 2016). The high spatial resolution of CCSM4-WRF allows distinguishing a stronger CRELWs over the eastern side of the Andes. The future changes in CRELWs are negative over the entire Andean domain, especially over the eastern high Andes in RCP8.5 (Fig. 13c). Figure 13d indicates that the CRELWs increases with elevation for both sides of the Andes, while its future changes tend to be more negative as elevation increases for both scenarios (Figs. 13e,f). While the scatter at low elevations in RCP4.5 influences the linear regression over the western slope, the elevation dependence toward more negative values at high altitudes is apparent in RCP8.5 for both sides of the Andes. These results indicate that the influence of clouds on net longwave radiation over the Andes is projected to decrease more strongly at higher elevation in the future. This pattern of CRELWs changes would actually tend to moderate the EDW signal, by preferentially decreasing QS at high elevations under climate change. However, the larger magnitude of the CRESWs changes results in a net contribution of the CRE to elevation-dependent warming.

Fig.  13.
Fig. 13.

As in Fig. 12, but for net longwave radiation at the surface (CRELWs).

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Figures 14a and 14d show scatterplots of the relationship between the projected changes in surface temperature (ΔT) versus the changes in CRE (ΔCRE) over the Andes. In general, ΔT increases with ΔCRE; that is, the less negative cloud radiative effect in the future at higher elevations (positive ΔCRE) is associated with stronger warming (higher ΔT).

Fig.  14.
Fig. 14.

Projected changes in (a) temperature (ΔT) vs projected changes in cloud radiative effect (ΔCRE), (b) projected changes of the cloud radiative effect (ΔCRE) against the projected changes in cloudiness (ΔCWP), and (c) projected changes in temperature (ΔT) against the projected changes in cloudiness (ΔCWP) for RCP4.5. (d)–(f) As in (a)–(c), but for the RCP8.5 scenario. Orange and green dots represent the western and eastern side respectively. The smaller black and white dots are plotted for clarity of the solid lines that represent the linear regression of each side, where the orange line represents the western and green line the eastern side. The solid black line represents the linear regression using all data from both sides of the Andes. The terms RW2 and RE2 are the coefficients of determination of the regressions for the western and eastern sides, respectively.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Figure 15a shows the cloudiness of the CTRL simulation by calculating the CWP defined in (2). In this simulation, the CWP is largest at high elevations on the eastern side. The projections of the future changes in the CWP (Figs. 15b,c) show an increase on the eastern slopes of the Andes between 500 and 2500 m, while the results are more mixed on the western slopes, even showing a decrease in RCP4.5. The higher elevations of the Andes, above 2500 m also experience a decrease in the CWP in both scenarios. This increase in cloudiness at lower elevations on the eastern slopes and the decrease above 2500 m is consistent with results in the previous section regarding future changes in the CRE and supports our hypothesis that the simulated EDW in the Andes of Ecuador may, at least in part, be a result of cloud-mediated changes in net surface radiation.

Fig.  15.
Fig. 15.

Changes in cloud water and ice mixing ratios as a proxy for cloudiness. The cloudiness is calculated by vertically integrating the cloud water mixing ratio and the ice water mixing ratio (cloud water path; see text). (a) Present-day (CTRL) cloudiness from CCSM4-WRF simulations. (b),(c) Changes in cloudiness with respect to CTRL for RCP4.5 and RCP8.5, respectively. Hatching indicates regions where cloud water path changes are significant at the 95% level based on a Student’s t test. The red contour lines represent the 500-m elevation, while the black solid contour lines indicate the elevation at intervals of 1 km, starting at 1500 m. The green solid contour lines indicate the 4000-m altitude. The yellow line indicates the latitude 0.07°N used for the vertical cross sections shown in Figs. 8 and 9.

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

Comparing the changes in CWP (ΔCWP) against ΔCRE for RCP4.5 and RCP8.5 scenarios shows that as ΔCWP becomes more negative, ΔCRE becomes more positive (Figs. 14b,e), reflecting the simulated future changes in the Andes above ∼2500 m, except for the interior valleys. This negative relationship results from the decrease in cloudiness (more negative ΔCWP), which leads to enhanced absorption of shortwave radiation at the surface. While the decrease in cloudiness also leads to a reduction in incoming longwave radiation at the surface, this reduction is weaker than the increase in shortwave radiation (hence a positive ΔCRE). Figures 14c and 14f show the influence that ΔCWP exerts on ΔT, with a decrease in cloudiness amplifying the warming signal seen in surface temperature. While there is significant scatter on the western slopes of the Andes in RCP4.5, the relationship is more evident on the eastern slope and in the RCP8.5 scenario.

e. The role of the snow-albedo feedback

Aside from cloud feedbacks, our results also suggest that at the highest peaks, reduced surface albedo due to loss of snow cover contributes to increases in warming rates. Figure 16a portrays the mean state of the present-day surface albedo simulated in CFSR-WRF over the highest terrain in the model. The highest peaks show the highest albedo (Fig. 16a) as a consequence of episodic snow cover simulated over high terrain (Fig. 16d). The future projections of surface albedo reveal changes over the highest peaks (Fig. 16b for RCP4.5, and Fig. 16c for RCP8.5). Over these locations snow cover is strongly diminished (Figs. 16e,f) in the future, leading to strong reductions in albedo. The reduced albedo leads to increases in absorbed solar radiation, QS (Fig. 10), and additional warming through the snow-albedo feedback.

Fig.  16.
Fig. 16.

Projected changes in albedo and snow cover from CCSM4-WRF simulations at high elevation. (a) Present-day (CTRL) albedo mean state. The black solid contour lines start at 1500-m altitude with intervals of 1 km, and the green solid contour line indicates the 4000-m altitude. (b),(c) Projected future changes in albedo for RCP4.5 and RCP8.5 scenarios respectively, when compared with the CTRL run. (d) Present-day (CTRL) snow cover mean state. (e),(f) Mean state of snow cover for the RCP4.5 and RCP8.5 scenarios, respectively. Note that color scale varies between (d), (e), and (f).

Citation: Journal of Climate 35, 3; 10.1175/JCLI-D-21-0379.1

4. Discussion

Observations suggest a discernible EDW effect on mean and maximum temperature in the Andes of Ecuador, although a clear detection is complicated by the lack of a dense station network along the Andes. Model simulations suggest that EDW will amplify in the future due to a number of feedbacks that operate on different spatial scales. While the EDW is muted in RCP4.5 as compared to RCP8.5, according to our simulations, the higher elevations in Ecuador will warm at enhanced rates in twenty-first century, regardless of emission scenario.

The enhanced large-scale warming seen in the tropical troposphere is likely a reflection of the enhanced release of latent heat during condensation associated with a convective adjustment of the tropical troposphere (Bony et al. 2006; Bradley et al. 2006). While diagnosing the exact reasons for this upper-tropospheric warming is beyond the scope of this study, it clearly projects onto surface temperature changes on both sides of the Andes, and is consistent with other modeling studies over this region (Bradley et al. 2006; Urrutia and Vuille 2009).

The differences we find in the EDW between eastern and western slopes have hitherto not been addressed in detail. Vuille et al. (2003) pointed out that cloud cover could have played a role in causing the differential warming between eastern and western slopes, but they did not analyze such a feedback. Our results suggest that changes in cloud cover could indeed provide an explanation of the differential warming of the eastern and western slopes seen in future projections. According to our simulations, changes in the zonal mean midtropospheric circulation lead to enhanced subsidence on the leeward side over the western slopes. This mechanism directly contributes to warming, but also does so indirectly, via inducing drying, reducing cloud cover and enhancing net surface radiation receipts at high elevations. Longwave and shortwave radiation are both affected by cloud cover changes, and the net radiative effect of clouds depends on their vertical distribution throughout the atmospheric column. Indeed, we found that changes in the CWP constitute a fundamental component of the cloud feedback mechanism, thereby contributing to the EDW effect over the Ecuadorian Andes. Our analyses also suggest that the CRE is dominated by the shortwave fluxes and that EDW is partially a consequence of the elevation-dependent changes of the CRESWs. Hence the future decrease in cloudiness is amplifying the warming signal seen in surface temperature at high elevation.

Finally, small areas at the highest elevations will also see enhanced warming due to the snow-albedo feedback, according to our simulations. This is consistent with a global analysis carried out by Pepin and Seidel (2005) and Pepin and Lundquist (2008), who showed that mean annual temperature trends are stronger in elevation bands from −5° to 5°C, close to elevations where snow/ice melting processes dominate. Their dataset relied primarily on locations in the Northern Hemisphere extratropics, with only a few stations from the tropics (and none from Ecuador), but our results suggest that this feedback may locally also play a role at high elevations in the tropics, similar to what has already been described in detail over the Sierra Nevada (Walton et al. 2017), Rocky Mountains (Minder et al. 2018), and European Alps (Kotlarski et al. 2015; Winter et al. 2017). The albedo feedback is strongest close to the annual mean 0°C isotherm, which tends to covary with the tropical glacier equilibrium line altitude (ELA) on interannual time scales (Vuille et al. 2018). The ELA (“snow line”) in the Andes of Ecuador currently varies on an interannual basis between 5000 and 5300 m, while the freezing line is located about 200 m below the snow line at about 4800–5100 m (Vuille et al. 2018). There is considerable interannual variability in both values due to ENSO, but monitoring the snow-albedo feedback and its contribution to EDW and enhanced glacier melt over Ecuadorian mountains in this broad elevation belt between ∼4800 and 5300 m MSL deserves further attention. Furthermore, the snow-albedo feedback also has the potential to induce nonlocal effects through regional scale circulation, thereby potentially also amplifying the warming over snow-free locations (Letcher and Minder 2015, 2017). The strength of the snow-albedo feedback in our simulations is subject to some uncertainty. For instance, the feedback strength is sensitive to the deposition of dust and aerosols on snow and ice, a process not represented in our simulations. While this aspect is still poorly understood in the tropical Andes (Molina et al. 2015), there is increasing evidence that black carbon deposition and biomass burning upstream over the Amazon basin may play a role in changing glacier albedo and enhancing melt rates on tropical Andean glaciers (Magalhães et al. 2019). Additionally, the simulated snow-albedo feedback, and its impact on EDW, can be strongly sensitive to the choice of land surface parameterization, especially representation of snow–vegetation–radiation interactions. Research over the Rocky Mountains suggests that the snow-albedo feedback and EDW may be too strong in simulations using the Noah land surface model, due to excessive fractional snow cover (Minder et al. 2016, 2018).

Finally, soil moisture is sometimes invoked to explain EDW over mountain ranges. This is reasonable, since soil moisture will be affected differently at different elevations, given the elevation dependence of projected future changes in precipitation and QS. Increased soil moisture can contribute to a greater absorption of incoming solar radiation at daytime, which will boost the outgoing longwave fluxes during nighttime, causing an increase in minimum temperature (Rangwala et al. 2012). Pepin et al. (2015) argued that maximum and minimum temperature will respond differently depending on soil moisture: if sensible heat fluxes balance the surface shortwave absorption, then this is reflected in higher maximum temperatures, while if latent heat fluxes are the balancing factor, minimum temperatures will increase. However, in our simulations, all run with the Unified Noah land surface model, we did not identify any significant elevation-dependent changes in soil moisture, regardless of emission scenario considered (not shown).

5. Summary and conclusions

Observational data suggest that there is a discernible EDW signal in mean and maximum temperature over the Ecuadorian Andes. Minimum temperature trends for the period 1986–2005, however, do not show a significant relationship with elevation. Due to the paucity of available data over the Andes, the detection of such elevation-dependent trends in observations is complicated and a clear attribution to a specific elevation-dependent forcing or feedback is not possible based on the limited observational data alone.

Therefore, four WRF experiments were used to investigate the EDW signal over the Ecuadorian Andes. CFSR-WRF and CCSM4-WRF CTRL both simulate an EDW signal over the region. This signal is amplified during the period from 2041 to 2060 in the two future scenarios RCP4.5 and RCP8.5.

Our simulations suggest that future EDW in the Andes of Ecuador results from a combination of different feedbacks, which include enhanced upper-tropospheric warming, reductions in high-elevation clouds that lead to increases in solar radiation receipts at the surface, and—over the highest peaks—a local snow-albedo feedback.

However, there are also clear and significant differences in the warming rates with elevation between the western and the eastern slope of the Andes. The western side is characterized by larger rates of warming with altitude: 0.08 and 0.12 K km−1 for RCP4.5 and RCP8.5 scenarios respectively, while the warming with altitude on the eastern side is 0.06 K km−1 in both scenarios. This difference may be caused by the increased subsidence on the western slopes due to changes in the mean zonal midtropospheric circulation. This increased subsidence can enhance warming directly by adiabatic descent, and it likely also contributes to changes in the CRE due to a reduced CWP. More detailed investigations into this effect at other tropical Andean sites seem warranted, to improve our understanding of how future changes in atmospheric circulation might affect cloud feedbacks and thus EDW and accelerated glacier retreat over the tropical Andes (e.g., Sicart et al. 2016). Unfortunately, the lack of in situ observations precludes a thorough assessment and validation of the modeled radiative fluxes at the surface. This is particularly problematic given that surface radiative fluxes appear to be key to understanding the mechanisms driving future EDW and its east–west asymmetry in the tropical Andes. Hence confirming the veracity of the magnitudes of the simulated radiation fluxes at the surface as simulated by our model should be a major component of future work.

Finally, a few inherent limitations associated with such modeling exercises should be kept in mind when interpreting the results from this study. For one, the resolution of regional climate models such as the one used here, while relatively high, is still not sufficient to explicitly represent convection without the use of subgrid parameterizations. Yet over regions that are strongly affected by convective storms, results from simulations with parameterized convection (∼10-km horizontal grid spacing) and those that explicitly simulate convection (∼1-km grid spacing) can exhibit substantial differences (e.g., Prein et al. 2015), which could affect both the cloud cover and the snow albedo feedback.

The results presented here are underpinned by only one regional model and based on boundary conditions from a single ensemble member of one global climate model (CSSM4). Therefore, additional studies on EDW in the tropical Andes using different model configurations to eventually create a multimodel ensemble would be highly welcome. The CORDEX-SA initiative already provides for such a framework (e.g., Solman and Blásquez 2019), but simulations at a higher resolution than what is provided in CORDEX-SA would be required to adequately resolve the terrain and address the mechanisms underpinning EDW in the Andes. Such an approach would help reduce these uncertainties, assuming that the ensemble mean yields a representation that is closer to observations (Flato et al. 2013). Nevertheless, it is worth keeping in mind that (i) we demonstrate that CCSM4 simulates future temperature and precipitation changes over the Ecuadorian Andes that are close to the CMIP5 multimodel mean and median (Fig. 2), (ii) the parameterizations used here have been tested successfully in prior studies over the tropical Andean domain (e.g., Moya-Álvarez et al. 2018; Martínez-Castro et al. 2019; Chimborazo and Vuille 2021), and (iii) WRF outperforms other RCMs over South America as shown in a recent CORDEX-SA analysis (Solman and Blásquez 2019).

Acknowledgments.

This study was funded by the U.S. State Department (S-LMAQM-11-GR-086) and the National Science Foundation (OISE-1743738, AGS-1349990). We thank three reviewers for the helpful comments and the Instituto Nacional de Meteorología e Hidrología del Ecuador (INAMHI) for providing station data. We acknowledge high-performance computing support from Yellowstone (ark:/85065/d7wd3xhc) provided by NCAR’s Computational and Information Systems Laboratory, sponsored by the National Science Foundation. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups for producing and making available their model output (listed in Table 3 of this paper). For CMIP the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. We acknowledge the groups that develop free software for analysis and plotting, including the python modules xarray (Hoyer and Hamman 2017), wrf-python (Ladwig 2017), Matplotlib (Hunter 2007), Cartopy (Met Office 2018), pyMannKendall (Hussain and Mahmud 2019); the NCL language (National Center for Atmospheric Research 2017), and the Climate Data Operators (CDO; Max Planck Institute for Meteorology 2018). The maps plotted with Cartopy used data from Natural Earth (free vector and raster map data at https://www.naturalearthdata.com/).

Data availability statement.

Output from WRF simulations is archived at the University at Albany, State University of New York and available upon request.

REFERENCES

  • Beniston, M., H. F. Diaz, and R. S. Bradley, 1997: Climatic change at high elevation sites: An overview. Climatic Change, 36, 233251, https://doi.org/10.1023/A:1005380714349.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bony, S. R., and Coauthors, 2006: How well do we understand and evaluate climate change feedback processes? J. Climate, 19, 34453482, https://doi.org/10.1175/JCLI3819.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Boucher, O., and Coauthors, 2013: Clouds and aerosols. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 571657.

    • Search Google Scholar
    • Export Citation
  • Bradley, R. S., M. Vuille, H. F. Diaz, and W. Vergara, 2006: Threats to water supplies in the tropical Andes. Science, 312, 17551756, https://doi.org/10.1126/science.1128087.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bradley, R. S., F. T. Keimig, H. F. Diaz, and D. R. Hardy, 2009: Recent changes in freezing level heights in the tropics with implications for the deglacierization of high mountain regions. Geophys. Res. Lett., 36, L17701, https://doi.org/10.1029/2009GL037712.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bruyère, C. L., A. J. Monaghan, D. F. Steinhoff, and D. Yates, 2015: Bias-corrected CMIP5 CESM data in WRF/MPAS intermediate file format. NCAR Tech. Note NCAR/TN-515+STR, 60 pp, https://doi.org/10.5065/D6445JJ7.

    • Search Google Scholar
    • Export Citation
  • Buytaert, W., M. Vuille, A. Dewulf, R. Urrutia, A. Karmalkar, and R. Célleri, 2010: Uncertainties in climate change projections and regional downscaling in the tropical Andes: Implications for water resources management. Hydrol. Earth Syst. Sci., 14, 12471258, https://doi.org/10.5194/hess-14-1247-2010.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Campozano, L., D. Ballari, M. Montenegro, and A. Aviles, 2020: Future meteorological droughts in Ecuador: Decreasing trends and associated spatio-temporal features derived from CMIP5 models. Front. Earth Sci., 8, 17, https://doi.org/10.3389/feart.2020.00017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carey, M., C. Huggel, J. Bury, C. Portocarrero, and W. Haeberli, 2012: An integrated socio-environmental framework for glacier hazard management and climate change adaptation: Lessons from Lake 513, Cordillera Blanca, Peru. Climatic Change, 112, 733767, https://doi.org/10.1007/s10584-011-0249-8.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cauvy-Fraunié, S., R. Espinosa, P. Andino, D. Jacobsen, and O. Dangles, 2015: Invertebrate metacommunity structure and dynamics in an Andean glacial stream network facing climate change. PLOS ONE, 10, e0136793, https://doi.org/10.1371/journal.pone.0136793.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569585, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, Y., C. M. Naud, I. Rangwala, C. C. Landry, and J. R. Miller, 2014: Comparison of the sensitivity of surface downward longwave radiation to changes in water vapor at two high elevation sites. Environ. Res. Lett., 9, 114015, https://doi.org/10.1088/1748-9326/9/11/114015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chimborazo, O., 2018: Projected changes in climate, elevation-dependent warming, and extreme events over continental Ecuador for the period 2041–2070. Ph.D. thesis, State University of New York at Albany, 227 pp., https://www.proquest.com/openview/4819b79e8b1f773235fef231dec73b8a/1?pq-origsite=gscholar&cbl=18750&diss=y.

    • Search Google Scholar
    • Export Citation
  • Chimborazo, O., and M. Vuille, 2021: Present-day climate and projected future temperature and precipitation changes in Ecuador. Theor. Appl. Climatol., 143, 15811597, https://doi.org/10.1007/s00704-020-03483-y.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Collins, M., and Coauthors, 2013: Long-term climate change: Projections, commitments and irreversibility. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 10291136.

    • Search Google Scholar
    • Export Citation
  • Cook, S. J., I. Kougkoulos, L. A. Edwards, J. Dortch, and D. Hoffmann, 2016: Glacier change and glacial lake outburst flood risk in the Bolivian Andes. Cryosphere, 10, 23992413, https://doi.org/10.5194/tc-10-2399-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dangles, O., A. Rabatel, M. Kraemer, G. Zeballos, A. Soruco, D. Jacobsen, and F. Anthelme, 2017: Ecosystem sentinels for climate change? Evidence of wetland cover changes over the last 30 years in the tropical Andes. PLOS ONE, 12, e0175814, https://doi.org/10.1371/journal.pone.0175814.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Drenkhan, F., C. Huggel, L. Guardamino, and W. Haeberli, 2019: Managing risks and future options from new lakes in the deglaciating Andes of Peru: The example of the Vilcabamba-Urubamba basin. Sci. Total Environ., 665, 465483, https://doi.org/10.1016/j.scitotenv.2019.02.070.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Eaton, B., 2011: User’s guide to the Community Atmosphere Model CAM-5.1. NCAR, http://www.cesm.ucar.edu/models/cesm1.0/cam/docs/ug5_1/ug.html.

    • Search Google Scholar
    • Export Citation
  • Feeley, K. J., and Coauthors, 2011: Upslope migration of Andean trees. J. Biogeogr., 38, 783791, https://doi.org/10.1111/j.1365-2699.2010.02444.x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Flato, G., and Coauthors, 2013: Evaluation of climate models. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 741866.

    • Search Google Scholar
    • Export Citation
  • Francou, B., M. Vuille, P. Wagnon, J. Mendoza, and J. E. Sicart, 2003: Tropical climate change recorded by a glacier in the central Andes during the last decades of the twentieth century: Chacaltaya, Bolivia, 16°S. J. Geophys. Res., 108, 4154, https://doi.org/10.1029/2002JD002959.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, Y., F. Chen, D. P. Lettenmaier, J. Xu, L. Xiao, and X. Li, 2018: Does elevation-dependent warming hold true above 5000 m elevation? Lessons from the Tibetan Plateau. npj Climate Atmos. Sci., 1, 19, https://doi.org/10.1038/s41612-018-0030-z.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Giorgi, F., J. W. Hurrell, M. R. Marinucci, and M. Beniston, 1997: Elevation dependency of the surface climate change signal: A model study. J. Climate, 10, 288296, https://doi.org/10.1175/1520-0442(1997)010<0288:EDOTSC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Goosse, H., 2015: Climate System Dynamics and Modelling. Cambridge University Press, 358 pp.

  • Harris, I., T. J. Osborn, P. D. Jones, and D. H. Lister, 2020: Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset. Sci. Data, 7, 109, https://doi.org/10.1038/s41597-020-0453-3.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hartmann, D. L., 2016: Global Physical Climatology. 2nd ed. Elsevier, 498 pp.

  • Heredia, M. B., C. Junquas, C. Prieur, and T. Condom, 2018: New statistical methods for precipitation bias correction applied to WRF model simulations in the Antisana region, Ecuador. J. Hydrometeor., 19, 20212040, https://doi.org/10.1175/JHM-D-18-0032.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hong, S., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341, https://doi.org/10.1175/MWR3199.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoyer, S., and J. Hamman, 2017: xarray: N-D labeled arrays and datasets in Python. J. Open Res. Software, 5, 10, https://doi.org/10.5334/jors.148.

  • Hu, X.-M., J. W. Nielsen-Gammon, and F. Zhang, 2010: Evaluation of three planetary boundary layer schemes in the WRF model. J. Appl. Meteor. Climatol., 49, 18311844, https://doi.org/10.1175/2010JAMC2432.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hunter, J. D., 2007: Matplotlib: A 2D graphics environment. Comput. Sci. Eng., 9, 9095, https://doi.org/10.1109/MCSE.2007.55.

  • Huss, M., and Coauthors, 2017: Towards mountains without permanent snow and ice. Earth’s Future, 5, 418435, https://doi.org/10.1002/2016EF000514.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hussain, M. M., and I. Mahmud, 2019: pyMannKendall: A Python package for non parametric Mann Kendall family of trend tests. J. Open Source Software, 4, 1556, https://doi.org/10.21105/joss.01556.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • IPCC, 2013: Climate Change 2013: The Physical Science Basis. T. F. Stocker et al., Eds., Cambridge University Press, 1535 pp., https://doi.org/10.1017/CBO9781107415324.

    • Search Google Scholar
    • Export Citation
  • Jacobsen, D., A. M. Milner, L. E. Brown, and O. Dangles, 2012: Biodiversity under threat in glacier-fed river systems. Nat. Climate Change, 2, 361364, https://doi.org/10.1038/nclimate1435.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Jiménez, P. A., J. Dudhia, J. F. González-Rouco, J. Navarro, J. P. Montávez, and E. García-Bustamante, 2012: A revised scheme for the WRF surface layer formulation. Mon. Wea. Rev., 140, 898918, https://doi.org/10.1175/MWR-D-11-00056.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Junquas, C., K. Takahashi, T. Condom, J. C. Espinoza, S. Chavez, J. E. Sicart, and T. Lebel, 2018: Understanding the influence of orography on the precipitation diurnal cycle and the associated atmospheric processes in the central Andes. Climate Dyn., 50, 39954017, https://doi.org/10.1007/s00382-017-3858-8.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kaser, G., M. Grosshauser, and B. Marzeion, 2010: Contribution potential of glaciers to water availability in different climate regimes. Proc. Natl. Acad. Sci. USA, 107, 20 22320 227, https://doi.org/10.1073/pnas.1008162107.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kotlarski, S., D. Lüthi, and C. Schär, 2015: The elevation dependency of 21st century European climate change: An RCM ensemble perspective. Int. J. Climatol., 35, 39023920, https://doi.org/10.1002/joc.4254.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ladwig, W., 2017: wrf-python (version 1.2.0). UCAR/NCAR, https://doi.org/10.5065/D6W094P1.