response of alpine glaciers in north ... - arpa piemonte · -700 m in the year 2090 super...
TRANSCRIPT
![Page 1: RESPONSE OF ALPINE GLACIERS IN NORTH ... - ARPA Piemonte · -700 m in the year 2090 SUPER ENSEMBLE-900 m in the year2090 theSRESA1Bscenario. Fig. 6: cumulated snout position in the](https://reader033.vdocuments.us/reader033/viewer/2022060702/606f58b970d5276b3e7520d3/html5/thumbnails/1.jpg)
RESPONSE OF ALPINE GLACIERS IN NORTH-WESTERN ITALIAN ALPS RESPONSE OF ALPINE GLACIERS IN NORTH-WESTERN ITALIAN ALPS FOR DIFFERENT CLIMATE CHANGE SCENARIOS
Riccardo BONANNO (1), Christian RONCHI (1), Barbara CAGNAZZI (1), Antonello PROVENZALE (2)
(1) Regional Agency for Environmental Protection - ARPA Piemonte, Torino - Italy(2) Institute of Atmospheric Sciences and Climate (ISAC) of the Italian National Research Council (CNR).
(1) Regional Agency for Environmental Protection - ARPA Piemonte, Torino - Italy(2) Institute of Atmospheric Sciences and Climate (ISAC) of the Italian National Research Council (CNR).
(corresponding author: [email protected])
ABSTRACTIn the densely populated Alpine regions, glaciers represent an important source of freshwater and a significant component of tourism economy and hydro-electric power production. The shrinking of glaciers inevitably leads to a reduction of the frozenwater supply they are able to store. This is one of the reasons why it is important to model and quantify, over time, the response of Alpine glaciers to different climate change scenarios.In this work, we analyzed the impact of climate variability on a set of glaciers in the north-western Italian Alps during the last 50 years, considering the fluctuations in glacier terminus position (or snout). The method adopted here involves the developmentand calibration of a linear empirical stochastic model, in which glacier snout variations depend on temperature and precipitation fluctuations. In the study of Calmanti et al., 2007, it has been shown that linear empirical models are able to reproduce andIn this work, we analyzed the impact of climate variability on a set of glaciers in the north-western Italian Alps during the last 50 years, considering the fluctuations in glacier terminus position (or snout). The method adopted here involves the developmentand calibration of a linear empirical stochastic model, in which glacier snout variations depend on temperature and precipitation fluctuations. In the study of Calmanti et al., 2007, it has been shown that linear empirical models are able to reproduce andpredict the mean response of glaciers to climate variability.The model is then used to estimate the average response of Alpine glaciers in different climate change scenarios, assuming that the selected predictors are suitable also for future climate conditions. The SRES A1B and the new RCPs scenario have beenconsidered. For all selected scenarios, the empirical model confirms the observed average retreat of glaciers in the NW Italian Alps during the whole XXI century. In the most dramatic case (SRES A1B regional scenario), the estimated retreat can be aslarge as one kilometer.large as one kilometer.
PURPOSE DATA
•Study of the average response of a set of glaciers in the north-western Italian Alps toclimate variability during the last 50 years and to future climate change scenarios.
•Method adopted: use of a lagged-linear empirical stochastic model, in which glacier snoutvariations depend on the fluctuations in temperature and precipitation.
Standardized average annual snout fluctuationsClimatic records of precipitation and temperature
in the period 1958-2009: Optimal Interpolation
variations depend on the fluctuations in temperature and precipitation.
AVAILABLE DATA•On the basis of theconsistent behavior ofthe different glaciers
• An Optimal Interpolation (OI) (Ronchi et al., 2008)technique is used to assimilate the daily ground station data,irregularly placed in the region, on a selected regular three-dimensional grid map based on a background field (BF).
2aAVAILABLE DATA
1. Annual fluctuations of glacier snout positions as a substitute for the quantitativemeasurement of glacier mass balance (proxy data).
2. Records of temperature and precipitation, available in the study area (north-western
the different glaciers(with a retreat which ismore or less pronounceddepending on themorphology of the
dimensional grid map based on a background field (BF).
• For temperature, the background field is obtained on aselected grid (0.125° resolution, with careful description ofthe complex orography of the region) by a linear tri-dimensional downscaling of ERA-40 archive from 1957 to
2a
2. Records of temperature and precipitation, available in the study area (north-westernItaly) in the period:
•1958-2009: observations distributed over a predefined regular grid (OptimalInterpolation)•2010-2100: coming from different climate change scenarios (RCPs and SRES A1B,see IPCC, 2007 and Moss et al., 2008.)
morphology of theglacier and the localmeteorology) weaveraged the normalizedtime series of thedifferent glaciers to
the complex orography of the region) by a linear tri-dimensional downscaling of ERA-40 archive from 1957 to2001 and of the ECMWF objective analysis from 2002 to2009.
see IPCC, 2007 and Moss et al., 2008.) different glaciers toproduce a signal whichdescribes the averageobserved behavior ofglaciers in the period1958-2009.
2b
1958-2009.
•The periods of retreatare alternated withperiods of advance, evenperiods of advance, evenif the overall effect isan average retreat ofabout 200 m throughoutthe period.
Fig 1: Schematic map of the study area (north-western Italy) : 14 glaciers considered
Fig. 3: The regular grid on which theOI has been calculated (the plotshows the orography in m).
Fig. 2 : a) Average standardized annual snout variation related to the period 1958-2009.b) Cumulated average snout variation in the same period. The resulting overall averageretreat from 1958 to 2009 is about 200 m.
METHODOLOGY
Fig 1: Schematic map of the study area (north-western Italy) : 14 glaciers considered retreat from 1958 to 2009 is about 200 m.
•From the precipitation and temperature data, we calculated the standardized monthly averages over the area that includes Piedmont and Valled'Aosta. The averages were then grouped in periods of varying duration, resulting in seasonal values. For each of these predictors, wecalculated the cross-correlation with the series of average snout fluctuations, determining for each variable the time delay (in years) thatmaximizes the correlation between the glacial snout and the meteorological variables. Snout fluctuations turn out to be positively correlatedwith precipitation and negatively with temperature.
Time series of monthly temperature and
precipitation from OI (Fig 3).Standardized and spatially
averaged values.
Cross-correlation Rx,y
To get the time lag for each
Aggregated seasonal variables with time lag
with precipitation and negatively with temperature.
•By screening the predictors with appropriate statistical techniques (such as backward stepwise regression and cross-validation for out ofsample prediction), we obtained a simple lagged-linear empirical stochastic model that is able to reproduce past snout fluctuations.
•The selected model depends only on four predictors: summer temperature and winter precipitation with time delay of five and ten yearsrespectively, spring temperature and precipitation in the year of the snout fluctuation. These predictors are meaningful in terms of glacier
3
averaged values.
Aggregated seasonal variable
ixδ
1
To get the time lag for each aggregated variable Y(n-m)
X = Y = T, PixδY∆t(n-m)
respectively, spring temperature and precipitation in the year of the snout fluctuation. These predictors are meaningful in terms of glacierphysics.
Spring temperature and precipitation: may change the surface glacier albedo in the summer ablation period Stochastic term
Backward stepwise regression: combination
of variables that minimizes the AIC
Aggregated seasonal variable
Y(1-3), Y(2-4), Y(3-5)….Y(1-4), Y(2-5), Y(3-6)….Y(1-5), Y(2-6), Y(3-7)….Y(1-6), Y(2-7), Y(3-8)….
2 Temperature
T(n-m)
Winter precipitation (accumulation period)
Summer temperature (ablation period)
minimizes the AIC (Akaike Information
Criterion).
Y(1-6), Y(2-7), Y(3-8)….
X = T, PY∆t(n-m)
Average snout fluctuation
2
4ixδ
Lag: response time of the average annual snout fluctuation at thetemperature / precipitation fluctuations of a certain period of the year
(accumulation period) (ablation period)
selection of thecombination that leads tothe best "out-of-sample“
Average annual snout fluctuation (Fig 2a)
4
Fig. 4 (on the left): Scheme of the technique used to select the predictors adopted in the model. For each predictor, we calculated the cross-correlation with the series of average snout fluctuations, determining for each variable the time delay (in years). Then, through a screening of thepredictors with appropriate statistical techniques, we obtained the simple lagged-linear empirical stochastic model illustrated above.
the best "out-of-sample“prediction.
fluctuation (Fig 2a)
ixδ
RESULTS
OUT-OF-SAMPLE deterministic prediction with the lagged-linear model
Impact of climate change on the glaciers of north-western Alps: RCPs and SRES A1B scenarios
Models used
• RCP scenarios: EC-EARTH GCM
• SRES A1B: outputs of 7 regional models
modelImpact of climate change on the glaciers of north-western Alps: RCPs and SRES A1B scenarios
•Good statistical score (R2 =0.93) and excellent out-of-
-500 m in the year 2090 -300 m in the year 2090
• SRES A1B: outputs of 7 regional modelsRCMs of the ENSEMBLES projectaggregated using the MULTIMODELSUPERENSEMBLE technique (Cane etal., 2012)
0.93) and excellent out-of-sample predictions show thatthe model is able to estimateglacier retreat in north-western Italy.
-500 m in the year 2090 -300 m in the year 2090al., 2012)
Results
We estimated the projections for thefuture evolution of the average snoutfluctuations in different climatescenarios. The most likely average
•We assume stationarity ofthe relationship betweenmeteorological predictorsand snout fluctuations (that
fluctuations in different climatescenarios. The most likely averagetendency for glaciers considered it is aretreat that in the most dramatic casescan be of the order of one kilometer. Themost critical situation is foreseen under
meteorological predictorsand snout fluctuations (thatis, we assume that the weightof the predictors in themodel does not change).
MULTIMODEL
SUPERENSEMBLE-700 m in the year 2090-900 m in the
year 2090
most critical situation is foreseen underthe SRES A1B scenario.
Fig. 6 : cumulated snout position in theperiod 1968-2088. The first part withwhite prediction bounds is the training
Fig. 5: ““““ out of sample ”prediction with the lagged-linear empirical model .
Conclusions and future developmentsCONCLUSIONS AND FUTURE DEVELOPMENTS
period 1968-2088. The first part withwhite prediction bounds is the trainingperiod 1958-2009.
Acknowledgements
prediction with the lagged-linear empirical model .
Conclusions and future developments
The aim of this work was to study the average behavior of a set of Alpine glaciers in the north-western Italian Alps in different climate changescenarios. We implemented a simple linear-lagged model that can be used to estimate the mean response of Alpine glaciers in different climatechange scenarios. The model is easy manageable as it depends only on four predictors, which can be justified in terms of glacier physics. The
CONCLUSIONS AND FUTURE DEVELOPMENTS AcknowledgementsWe acknowledge useful discussions with Nicola Loglisci and Renata Pelosini of ARPA Piemonte. We also thank Jost vonHardenberg for providing us with the data of the RCPs scenarios obtained with the EC-Earth model.
References
• Calmanti S., Motta L, Turco M. and Provanzale A. 2007. Impact of climate variability on Alpine glaciers inchange scenarios. The model is easy manageable as it depends only on four predictors, which can be justified in terms of glacier physics. Thecross-validation shows that the model is reliable for estimating glacier retreat in north-western Italy in the near future, assuming stationarityof the relationship between meteorological predictors and snout fluctuations.Using this model, we estimated the future behavior of the average glacier snout fluctuations in different climate scenarios. All scenarios weinvestigated show a common trend of glacier retreat in the north-western Italian Alps. The rate of glacier retreat depends on the scenario, withan acceleration starting around 2050 for the RCP 8.5 and SRES A1B scenarios. In the most dramatic case (A1B scenario) the average retreat of
• Calmanti S., Motta L, Turco M. and Provanzale A. 2007. Impact of climate variability on Alpine glaciers innorthwestern Italy. International Journal of Climatology 27, 2041–2053
• Cane D., Barbarino S., Renier L. A., and Ronchi C., “Regional climate models downscaling in the Alpine area withMultimodel SuperEnsemble”, Hydrol. Earth Syst. Sci. Discuss., 9, 9425-9454, 2012.
• Comitato Glaciologico Italiano, 1959 : Catasto dei Ghiacciai Italiani (Inventory of Italian Glaciers), 1, Elencogenerale e bi bliografia dei ghiacciai italiani.
an acceleration starting around 2050 for the RCP 8.5 and SRES A1B scenarios. In the most dramatic case (A1B scenario) the average retreat ofthe glaciers is as large as one kilometer.As a word of caution, we note that for strong glaciers retreat the assumption of linear response to climate forcing may be invalid. In addition,while the simulations with the RCPs scenarios are generated by a GCM, the A1B scenario simulation comes from a blend of regional models: thismay explain some of the differences in the results obtained with the A1B and the RCP scenarios.
generale e bi bliografia dei ghiacciai italiani.
• Comitato Glaciologico Italiano, 1961 : Catasto dei Ghiacciai Italiani (Inventory of Italian Glaciers), 2, Ghiacciai delPiemonte.
• Ronchi C., De Luigi C., Ciccarelli N., Loglisci N. 2008. Development of a daily gridded climatological air temperaturedataset based on a optimal interpolation of ERA-40 reanalysis downscaling and a local high resolution thermometersnetwork. Poster presentation at 8th EMS Annual Meeting & 7th European Conference on Applied Climatology, 29September–3 October 2008, Amsterdam, The Netherlands.may explain some of the differences in the results obtained with the A1B and the RCP scenarios. September–3 October 2008, Amsterdam, The Netherlands.