Академический Документы
Профессиональный Документы
Культура Документы
2
I. Historical Climate Background
A. Climate in the Tropics
1. Seasonal Cycle near the Equator
3
The El Niño events of 1982-1983 and 1997-1998 were particularly strong
in regards to impacts for Ecuador. La Niña conditions have occurred less
often over the past 35 years than El Niño conditions (Figure 1), and
generally do not have as significant effect for equatorial South America
(though effects elsewhere may be notable).
Figure 1: Multivariate ENSO index shows El Niño (red) and La Niña (blue) phases.
Six variables are used to determine the index: sea level pressure, east-west and
north-south components of the surface wind, sea surface temperature, surface air
temperature, and total amount of cloudiness. [Source: Climate Diagnostics Center,
NOAA/CIRES, http://www.esrl.noaa.gov/psd]
The range of impacts on Ecuador and the Ecuadorian Andes from ENSO is
not well-understood. One recent study suggests that the influence of
ENSO on the Ecuadorian Andes is limited, with significant precipitation
influence ending at the western edge of the mountains (Figure 2) [Rossel
and Cadier, 2009]. On the other hand, researchers examining glacier
behavior on Antisana Volcano in the northern Ecuadorian Andes found a
connection between El Niño /La Niña phases and glacier mass balance
(see section IC) [Francou, et al., 2004]. Discrepancies such as these point
toward the wide range of scientific opinion in regards to ENSO and its
Ecuadorian impacts. Nevertheless, ENSO is likely to remain an important
factor in Ecuador’s climate as the global climate changes.
i. Predicting ENSO
4
Figure 2: Boundaries of strong (>40%) significant (>20%) ENSO influences on annual
rainfalls in northwestern South America [Rossel and Cadier, 2009].
B. Climate in Ecuador
1. Overview of Ecuador’s Climatology
5
(hereafter referred to as UDel data) . These gridded data were produced
1
i. Temperature in Ecuador
The UDel data is only available over land, whereas the model outputs may
incorporate some fraction of output over the ocean. This potential
source of bias will be discussed in Section IIA. Maps of the average
temperature and precipitation over the whole time series are shown in
Figure 4 and Figure 5.
Figure 3: The grid box used to extract UDel data (only sub-grid boxes over land were
used).
1
The data is provided in ASCII format from the University of Delaware:
http://climate.geog.udel.edu/~climate/. We were provided with a gridded netCDF format version by
University of Washington graduate student Rob Nicholas.
2
Version 2.01 of the 1900 – 2008 temperature product; Matsuura and Willmott, 2009.
3
Version 2.01 of the 1900 – 2008 precipitation product; Matsuura and Willmott, 2009.
6
Figure 4: Temperature from the UDel dataset averaged over the entire time series
(1900-2008). Grey line delineates the coast, and dark black contours delineates
topography. Blue stars show the locations of selected weather stations. Scale in oC.
7
Figure 6: Monthly temperature data (UDel), showing a linear warming trend over the
last century of about 0.1o/decade.
8
Figure 7: The frequency of months at, above, or below the mean temperature
(normalized to 0oC to emphasize deviation from the mean) for the grid box.
Figure 8: As in Figure 7, divided up by season. Note how the distribution differs for
each season.
9
Figure 9: Historical seasonal cycle of monthly mean temperature, which varies by
only a few degrees due to proximity to the equator.
Figure 10: The historical precipitation record (UDel) for 1900-2008 shows no clear
trend.
10
Figure 11: The seasonal cycle of precipitation for the entire grid box. The wettest
month is March and the driest is August.
11
anomalies, based on the regions of the tropical Pacific shown in Figure
13, and accessed from the National Weather Service4.
Figure 13 The sea surface temperature indices of ENSO that we consider are based
on the areas of ocean shown in the map. Niño1+2 is best correlated with
precipitation in Ecuador. (http://www.bom.gov.au/climate/enso/indices/oceanic-
indices-map.gif)
It results that the SST indices for the combined region of Niño1+2
correlate the best with precipitation anomalies, unsurprising since the
Niño1+2 regions are just off the coast of Ecuador. The correlations are
significant (to 95% confidence) in the months in which precipitation is
greatest (November-May) and the correlation is strongest (and not well
correlated or significant in the other months – June to October).
The SST index that best predicts ENSO, however, is Niño 3.4. Because
Niño 3.4 is removed from the Ecuadorian coast, it is clear that Ecuadorian
precipitation and temperature anomalies are not influencing the SST
index, so it better isolates the extent to which anomalies in Ecuador are
related to ENSO. The results of the correlations with each of the SST
indices are presented in Table 1.
The time series plots of Niño 3.4 anomalies and monthly temperature
anomalies are shown in Figure 14Error! Reference source not found..
Precipitation anomalies are compared with Niño 3.4 in Figure 15. The
best-correlated month (April), which is also one of the rainiest months, is
shown by itself in Figure 16Error! Reference source not found.. This
suggests that based on historical data, and baring any drastic change in
the state of ENSO in the future, indices of the strength of ENSO do a good
job of indicating when precipitation in the rainy months is particularly
high.
4
http//www.cpc.ncep.noaa.gov/data/indices/
12
Table 1: Correlations of SST index anomalies with monthly precipitation anomalies
from UDel data. Correlations that are significant to the 95% level are in bold.
Figure 14: Time series of the relative NINO 3.4 anomaly and the relative surface
temperature anomaly averaged over all of Ecuador.
Figure 15: Time series of the relative NINO 3.4 anomaly and the relative precipitation
anomaly averaged over all of Ecuador.
13
Precipitation and SST Anomalies for April
6
Precip Anomalies
5 Niño 3.4 Anomalies
3
Anomaly
2
−1
−2
−3
1950 1960 1970 1980 1990 2000
Year
Figure 16: Time series of precipitation anomalies from UDel data – April only,
compared with SST anomalies of Niño1+2 indices.
Figure 17: Bar graph showing the correlation coefficient between the NINO 3.4 index
anomaly and the precipitation anomaly averaged over all of Ecuador. Values that are
statistically significant to 95% are denoted with * and values that are statistically
significant to 99% are denoted with **.
14
Figure 18: Bar graph showing the correlation coefficient between the NINO 3.4 index
anomaly and the temperature anomaly averaged over all of Ecuador. Values that are
statistically significant to 95% are denoted with * and values that are statistically
significant to 99% are denoted with **.
The discussion in Section IA2 indicates that the influence of ENSO should
be looked at spatially (Figure 2). Below we compare the correlation
between ENSO, precipitation, and temperature for the coast and for the
sierra. For precipitation (Figure 19 and Figure 21) the correlations
improve slightly for the coast and deteriorate for the sierra region. For
temperature (Figure 20 and Figure 22) the correlations do not change
(r<0.05) for the coast and decrease for the sierra.
The monthly correlations for ENSO and precipitation and ENSO and
temperature are shown in Figures 19-20 for Ecuador’s coast.
15
Figure 19: Bar graph showing the correlation coefficient between the NINO 3.4 index
anomaly and the precipitation anomaly averaged over the coastal region. Values that
are statistically significant to 95% are denoted with * and values that are statistically
significant to 99% are denoted with **.
Figure 20: Bar graph showing the correlation coefficient between the NINO 3.4 index
anomaly and the temperature anomaly averaged over the coastal region. Values that
are statistically significant to 95% are denoted with * and values that are statistically
significant to 99% are denoted with **.
16
The monthly correlations for ENSO and precipitation and ENSO and
temperature are shown in Figures 21-22 for the Sierra.
Figure 21. Bar graph showing the correlation coefficient between the NINO 3.4 index
anomaly and the precipitation anomaly averaged over the sierra region. Values that
are statistically significant to 95% are denoted with * and values that are statistically
significant to 99% are denoted with **.
Figure 22. Bar graph showing the correlation coefficient between the NINO 3.4 index
anomaly and the temperature anomaly averaged over the sierra region. Values that
are statistically significant to 95% are denoted with * and values that are statistically
significant to 99% are denoted with **.
17
2. Regional Climates within Ecuador
18
Figure 24. Mean annual temperature cycle for Quito (Sierra), Guayaquil (Coast), Puyo
(Amazon), and San Cristóbal (Pacific Island) prior to 2000. [Bermeo et al., 2000]
Figure 25. Time series is an example of the diurnal temperature cycle in Quito and
Guayaquil, Ecuador (from April 23 through April 24, 2010). Relative to the annual
cycle, the changes in temperature in the diurnal cycle are much larger. (Weather
Underground).
19
Figure 26. Mean annual precipitation cycle for Quito (Sierra), Quayquil (Coast), Puyo
(Amazon), and San Cristóbal (Pacific Island) prior to 2000. [Bermeo et al., 2000]
The precipitation cycle for different regions of Ecuador is identified in Table 2
C. Changing Glaciers
1. Andean Glaciers
The Andes hold more than 99% of all tropical glaciers, with Peru holding
~70% of those and the other significant glaciers locations being primarily
in Ecuador and Bolivia. With icemelting throughout the year,
communities throughout the Andes rely on glaciers for some or all of
their seasonal water supplies (especially during dry seasons). The future
of Andean glaciers, however, is not bright. Evidence for past, present,
and future loss of glaciers throughout the Andes is present through
direct observations of mass balance [e.g., Francou et al., 2003; Soruco, et
al., 2009] and indirect indications from studies examining related factors,
. In the coastal region, most of the rainfall is along the northern portion
of the country (Figure 26). Along the Andes, rainfall becomes scarcer with
altitude [Bermeo et al., 2000]. The Amazon has permanent rainfall
throughout the year and is Ecuador’s rainiest region.
Table 2. Rainy seasons and peak rainfall months for different regions of Ecuador.
[Bermeo et al. 2000].
20
C. Changing Glaciers
1. Andean Glaciers
The Andes hold more than 99% of all tropical glaciers, with Peru holding
~70% of those and the other significant glaciers locations being primarily
in Ecuador and Bolivia. With icemelting throughout the year,
communities throughout the Andes rely on glaciers for some or all of
their seasonal water supplies (especially during dry seasons). The future
of Andean glaciers, however, is not bright. Evidence for past, present,
and future loss of glaciers throughout the Andes is present through
direct observations of mass balance [e.g., Francou et al., 2003; Soruco, et
al., 2009] and indirect indications from studies examining related factors,
Figure 26. Figure demonstrates the topography and rainfall throughout Ecuador.
There is relatively little precipitation along the Sierra and southern coast, but more
rainfall in the Amazon and along the northern coast. (Ecominga, 2010).
2. Ecuador’s Glaciers
Roughly speaking, Ecuador’s mountains and, thus, its glaciers, lie within
two mountain ranges, the Cordillera Occidental (more western peaks) and
the Cordillera Oriental (more eastern peaks, bordering the Amazon).
Glaciers in the Cordillera Oriental are found at lower elevation because of
21
the high peaks on which they sit and the high precipitation brought in
from the Amazon basin and tropical Atlantic. Ecuador’s glaciers have
already seen important changes that are linked to climate change.
Cotacachi, Corazon, and Sincholagua peaks, in the Cordillera Occidental,
completely lost their permanent ice in the last 10-15 years, while the
higher peaks of the Cordillera Oriental (Chimborazi, Cayambe, Antisana,
and Cotopaxi) are continuing to experience ice retreat (Figure 29).
Figure 27: Change in length and surface area of 10 tropical Andean glaciers from
Ecuador (Antisana 15a and 15b), Peru (Yanamarey, Broggi, Pastoruri, Uruashraju,
Gajap) and Bolivia (Zongo, Charquini, Chacaltaya) between 1930 and 2005. [Vuille, et
al., 2008]
22
Figure 28: Satellite image showing Quito (red star) and Antisana Volcano, Cayambe
Volcano, and Cotopaxi Volcano (blue circles). [Image source: Google Earth]
El Niño
o With El Niño events, the Andes are likely to experience warm
and dry anomalies
Higher air temperatures favor rainfall over snowfall
Glacier albedo (the amount of radiation reflected) is
decreased and the ice absorbs more radiation
Generally low wind speeds, which encourage melting
Less cloud cover, which allows for higher incoming
radiation
La Niña
o The Andes experience cold and wet weather during La Niña
events.
Cold temperatures encourage snowfall over rainfall
Increased snowfall helps to maintain high albedo,
reflecting more radiation
24
Generally higher wind speed that help reduce melting
Figure 30 Antizana mass balance anomalies (mm water equivalent) in the ablation
zone stratified by ENSO events. Individual monthly measurements (small circles),
the mean (large circles), and +/- 1 standard deviation are indicated (vertical bars) for
El Niño (solid circles) and La Niña events (open circles). [Francou, et al., 2004]
25
- “ "In 20 to 30 years we will have a problem with the potable water
supply," says Bolivar Caceres, a glaciologist with the hydrology and
meteorology institute. As the glaciers recede, he says, there will be
less water for Quito, where 70 percent of the water comes from
surrounding ice caps.“ [Bartolone, 2006]
- About 10 per cent of the water supply to Quito, the capital of
Ecuador, is also estimated to come from surrounding ice caps.”
[Painter, 2007]
26
Figure 31 Models included in the IPCC AR4, and their IDs with the Coupled Model
Intercomparison Project.
Climate model predictive skill for future climates depends on the area
over which the projection is being made and the variable in question.
Global circulation models can replicate historical temperature trends, but
precipitation is not as easily represented. Even though climate models are
able to replicate historical climates, there are still large uncertainties in
predicting future climate conditions both because the future
anthropogenic influence is unknown and because it is unclear how large
climate feedbacks will be (a climate feedback is something that enhances
or diminishes an effect of climate change).
27
i. Climate Model Predictive Skill for Temperature
Global climate models have been able to reproduce historical 20th century
temperature trends when known climate forcings or influences are taken
into account (Figure 32). While this report does not attempt to
demonstrate good agreement between historical Ecuadorian temperature
trends and climate model predictions, the mean model bias for surface-
air temperature in this region of South America is typically less than 2oC
on land (Figure 33). Temperature biases may be a result of issues such as
ocean contamination in land dominated grid boxes or altitudinal effects
(flattening a grid box when there is topography). Despite temperature
biases, the temperature trend (temperature change over time) is expected
to be representative of future climates, even if the predicted trend is not
well constrained (Figure 34).
28
Figure 33. MMD ensemble annual mean surface air temperatures in South America
compared with observations. a) observations from the HadCRUT2v data set (Jones et
al., 2001); b) mean of the 21 MMD models; c) difference between the multi-model
mean and the Had- CRUT2v data [IPCC, 2007].
29
Figure 34. Temperature anomalies with respect to 1901 to 1950 for the Amazon
region of South America for 1906 to 2005 (black line) and as simulated (red
envelope) by MMD models incorporating known forcings; and as projected for 2001
to 2100 by MMD models for the A1B scenario (orange envelope). The bars at the end
of the orange envelope represent the range of projected changes for 2091 to 2100 for
the B1 scenario (blue), the A1B scenario (orange) and the A2 scenario (red). The black
line is dashed where observations are present for less than 50% of the area in the
decade concerned [IPCC, 2007].
30
Figure 35. MMD ensemble annual mean precipitaition in South America compared
with observations. a) Observations (CMAP) are an update of Xie and Arkin (1997).
Units mm/day); b) mean of the 21 MMD models; c) difference between the multi-
model mean and the CMAP data [IPCC, 2007].
31
Figure 36. Temperature and precipitation changes over Central and South America
from the MMD-A1B simulations. Top row: Annual mean, DJF and JJA temperature
change between 1980 to 1999 and 2080 to 2099, averaged over 21 models. Middle
row: same as top, but for fractional change in precipitation. Bottom row: number of
models out of 21 that project increases in precipitation [IPCC, 2007].
32
1. Dynamical Downscaling
2. Statistical Downscaling
Figure 38: A schematic of the energy budget for glaciers. While many of these terms
are not discussed, the take-home message is the complexity of the picture
(http://geosci.uchicago.edu/~rtp1/glaciers/EnergyBudget.html)
34
III. Future Climate Predictions for Ecuador
A. Future Temperature and Variability
1. Predicting Future Temperature
In cases where distinctions were made between the coast and the Andes,
the coast was defined as areas west of 79o W longitude and the sierra was
defined as 77o W to 77.5o W longitude.
35
Table 3 The IPCC model output used in the analysis of future trends.
To determine the future warming for Ecuador, the warming trend was
calculated from the difference between the average of the historical
36
scenario for each climate model (1980 – 2000) and the future scenario for
each model (2030 – 2050 and 2080 – 2100) (Figure 39 and Figure ).
Figure 39:. Linear trend for our Ecuador GCM grid cell during the period from 1990 –
2040 for each model used in this report. The mean trend is 0.23oC decade-1.
Figure 40: Linear trend for our Ecuador GCM grid cell during the period from 1990 –
2090 for each model used in this report. The mean trend is 0.31oC decade-1.
The probability distribution function from data (1900 – 2008) was then
37
shifted by the calculated model temperature change (for both 2030 –
2050 and 2080 – 2100). All 19 model probability distribution functions
were averaged to produce future probability distribution functions (Figure
and Figure ). The total warming from all models were averaged together
and this value was added to the historical temperature data for each
spatial point (ignoring lapse rate changes) (Figure 37 and Figure 38).
Probability Distribution of Temperature: 2040
35
Historical
Model
30
25
20
15
10
0
18 20 22 24 26 28 30
o
C
Figure 41: Probability distribution functions for monthly historical de-trended
Ecuadorian temperatures (1900 – 2008) about a mean value of 22.8oC and the shifted
probability distribution function for the future climate (2030 – 2050).
25
20
15
10
0
18 20 22 24 26 28 30
o
C
Figure 42. Probability distribution functions for monthly historical de-trended
Ecuadorian temperatures (1900 – 2008, blue) about a mean value of 22.8oC and the
shifted probability distribution function for the future climate (2080 – 2100, red).
38
Figure 37. Historical spatial temperature distribution (1980 – 2000 – left figure) and
the shifted spatial temperature distribution with the mean temperature increase from
model output added to each grid box (1.13oC) for 2030 – 2050 (right figure).
Figure 38. Historical spatial temperature distribution (1980 – 2000 – left figure) and
the shifted spatial temperature distribution with the mean temperature increase from
model output added to each grid box (3.13oC) for 2080 – 2100 (right figure).
Similarly, this approach was taken for each seasonal cycle. In this case
the seasons were averaged (December – February, March – May, June –
August, September – November) and the probability distribution function
was calculated for the historical data for each season. The model
calculated change for each season was then used to shift the probability
distribution function. The shifted probability distribution functions were
then averaged together to form the future seasonal scenarios (Figure 39
and Figure 40).
39
Temperature PDF by Season: 2040
March − May June − Aug
30 30
20 20
10 10
0 0
20 25 30 20 25 30
30 30
20 20
10 10
0 0
20 25 30 20 25 30
30 30
20 20
10 10
0 0
20 25 30 20 25 30
30 30
20 20
10 10
0 0
20 25 30 20 25 30
40
Table 4 Historic and future temperature trends for Ecuador.
Note in Table 4 that temperature trends are typically greater in the sierra
than along the coast (historically). This phenomenon of greater warming
at altitude is predicted into the future.
Figure 47. Global warming in the American Cordillera. Projected changes in mean
annual free-air temperatures between (1990 to 1999) and (2090 to 2099) along a
transect from Alaska (68°N) to southern Chile (50°S), following the axis of the
American Cordillera mountain chain. Results are the mean of eight different general
circulation models used in the 4th assessment of the Intergovernmental Panel on
Climate Change (IPCC) (15), using CO2 levels from scenario A2 in (16). Black triangles
denote the highest mountains at each latitude; areas blocked in white have no data
(surface or below in the models). [Bradley et al, 2006].
Unfortunately, the instances of frost were not taken directly from general
circulation model output for this report and the future of frost incidences
in Ecuador is uncertain. Radiative cooling at night causes the majority of
frost in the Andes (Trognitz) and in some studies frost in the Andes was
associated with freezing air temperatures only 22% of the time [Villegas,
1991]. Since direct station data was not analyzed on a daily basis, frost
trends are not directly measured in this report.
41
For radiative cooling to occur to this degree, clear-sky conditions must
exist. Again, this report did not look directly at the cloud-fraction from
climate model predictions, but there is a useful analog between
precipitation and cloud cover. Richards and Arkin (1980) showed that for
large-scale (several degrees latitude) gridboxes precipitation and cloud-
cover are well correlated (r>0.8 in many cases). With our expectation for
increased rainfall throughout the year (Section III B), this suggests that
there will be increased cloudiness and more insolation to reduce the
number of frost days. Further, correlating temperature anomalies and
precipitation anomalies since 1950 show statistically significant and
positive correlations for April, May, June, and July (p<0.05). This is
further anecdotal evidence that less frost should be expected in the
future, since precipitation is expected to increase. Importantly, though,
monthly temperature anomalies are not a good representation of the
daily diurnal cycle, especially on regional scales.
42
Figure 41. Figure shows the average annual precipitation cycle for both historical
data and each model over the period from 1950 – 2000.
43
Figure 42. Figure shows the average annual precipitation cycle for both historical
data and the best fit models (r > 0.7, rms error < 3) over the period from 1950 –
2000.
44
Figure 50: Probability distribution functions for monthly historical Ecuadorian
precipitation (1900 – 2008) and the shifted probability distribution function for the
future climate (average for 2030 – 2050).
45
one model predicts a doubling of precipitation, while the other two do
not. Such discrepancies serve as a reminder not to over-interpret these
results.
46
In summary, the predicted shifts in temperature and precipitation by
2040 and 2090 are compiled in Table 5.
47
Figure 43. Figure shows the 2040 temperature increases expressed in terms of an
ENSO anomaly (NINO 3.4) for each month. The error is shaded red and is derived
from the fit error between historical temperature anomalies and ENSO anomalies.
The standard deviations for the mean ENSO state are also plotted for reference.
Figure 44. Figure shows the 2090 temperature increases expressed in terms of an
ENSO anomaly (NINO 3.4) for each month. The error is shaded red and is derived
from the fit error between historical temperature anomalies and ENSO anomalies.
The standard deviations for the mean ENSO state are also plotted for reference.
48
similarly be explored in terms of ENSO state (Figure 45 and Figure 46).
Figure 46: Figure shows the 2090 precipitation changes expressed in terms of an
ENSO anomaly (NINO 3.4) for each month. The error is shaded red and is derived
from the fit error between historical precipitation anomalies and ENSO anomalies.
The standard deviations for the mean ENSO state are also plotted for reference. Only
months denoted with * have a statistically significant relationship between ENSO
anomaly and precipitation anomaly (p<0.05).
49
Further, the correlations are not high enough to make a confident
comparison between future climate and ENSO anomaly for every month.
Figure 47. Plot shows the historical precipitation for two large ENSO anomalies, the
mean historical precipitation cycle, and the predicted precipitation for the 2040
climate.
50
Figure 48. Plot shows the historical precipitation for two large ENSO anomalies, the
mean historical precipitation cycle, and the predicted precipitation for the 2090
climate.
Figure 60: Plot shows the historical temperature for two large ENSO anomalies, the
mean historical temperature cycle, and the predicted temperature for the 2040
climate.
51
Figure 49:. Plot shows the historical temperature for two large ENSO anomalies, the
mean historical temperature cycle, and the predicted temperature for the 2090
climate.
As a final comparison, the actual ENSO indices for the 1997-1998 and
1982-1983 El Niño are compared with expectations for future
precipitation and temperature (expressed in terms of ENSO anomalies).
52
Figure 51. Future temperature changes expressed in terms of ENSO anomalies
(months that have a statistically significant correlation between precipitation
anomaly and ENSO anomaly are denoted with *). By 2040 the temperature changes
are comparable to a one standard deviation increase in ENSO state and by 2090 the
typical climate (in terms of temperature) will be greater than the large El Niño events
of 1982-1983 and 1997-1998. Error bars have been omitted, but are available in
Figure 43 and Figure 44.
53
IV. References
Bartolone, P. (2006). When the water runs out, Salon, Web, April 20, 2010,
< http://www.salon.com/news/feature/2006/04/07/ecuador2>
Bermeo, A., Cáeres-Silva, L., and Saari, P. (2000), eds., Climate Change,
National Communication: Republic of Ecuador.
Bradley, Raymond S., Vuille, Mathias, Diaz, Henry F., and Vergara, Walter
(2006), CLIMATE CHANGE: Threats to Water Supplies in the Tropical
Andes, Science, 312 (5781), doi: 10.1126/science.1128087.
Francou B., M. Vuille, V. Favier, and B. Caceres (2004). New evidence for an
ENSO impact on low-latitude glaciers: Antizana 15, Andes of Ecuador, 0
degrees 28 ' S, J. Geophys. Res., 109 (D18), doi: 10.1029/2003JD004484.
54
Geographic Guide (2010), Ecuador Map, Geographic Guide, Web, 30 May
2010, <http://www.geographicguide.net/america/ecuador.htm>
55
961.
Roe, G.H. and M.A. O’Neal, (2009). The response of glaciers to intrinsic
climate variability: observations and models of late Holocene variations.
J. Glaciology, 55, 839-854.
56
Physics and Meteorology, 141 pp.
57