The Use of Computer Modelling for Surface Mass Balance Analysis
Horst Machguth | Natural History Museum of Denmark, Copenhagen, and WGMS, Zurich, Switzerland
Based on the work of Matthias Huss and Leo Sold, University of Fribourg and ETH Zurich, Switzerland Use of Computer Modelling Introduction | 1
Goal • Understand advantages and disadvantages of computer modelling compared to traditional concepts. • Highlight the importance of comprehensive field observations
Use of Computer Modelling Introduction | 2
Overview
1. The traditional way (surface mass balance from the contour line method) 2. Surface mass balance from computer modelling 3. Conclusions – advantages and disadvantages of the approaches
Use of Computer Modelling introductionIntroduction | 3 Example Findel glacier | basic glacier characteristics
• Mid-sized (13 km2) photo: G. Kappenberger valley glacier in the Swiss Alps • Mass balance observations starting October 2004
Use of Computer Modelling Introduction | 4 Example Findel glacier | glaciological measurements 2004 - present
• ~12 stakes • Extensive snow probings in spring and radar measurements of snow thickness • Furthermore weather stations Zermatt (1640 m a.s.l., 6 km distance) and Gornergrat (3110 m a.s.l., 4 km distance)
Use of Computer Modelling Traditional approach | 5
A brief recapitulation of the Contourline Method
Use of Computer Modelling Traditional approach | 6 Contour Line Method | annual surface mass balance Findel year B (m w.e.) s (uncer- tainty) 2005/06 -1.2 0.5 • Contour line 2006/07 -0.2 0.5 method, manual 2007/08 -0.3 0.5 surface mass 2008/09 0.1 0.5 balance analysis 2009/10 -0.47 0.5 2005-10 -2.07 1.1 Use of Computer Modelling Traditional approach | 7 Contour Line Method | using expert knowledge
Expert knowledge 1: GPR flight May 2005 indicates very large amounts of accumulation close to the main crest
Expert knowledge 2: taking exposition into account
Expert knowledge 3: low accumulation on wind exposed crest Use of Computer Modelling Traditional approach | 8 Contour Line Method | Limitations
• Difficult to incorporate all available data • Limited possibilities to extrapolate data in space • Almost impossible to interpolate in time (i.e. confined to the temporal resolution of the measurements) • Difficult to reproduce Use of Computer Modelling Computer Modelling | 9
Computer modelling
• Basics • Implementation • Results • Large scale application
Use of Computer Modelling outlines of observationalComputer Modelling network | 10 Basics | Mass and Energy Balance at the glacier surface
? ... rather than two discrete measurements, knowing the temporal evolution would be valuable ... source: modified after M.Huss Use of Computer Modelling outlines of observationalComputer Modelling network | 11 Basics | the concept
1. Numerically simulate the mass and energy balance to extrapolate point measurements in space and time (Already applied by e.g. Braithwaite (1983) for temporal extrapolation of individual stake readings, but the approach demonstrated in the following is much more comprehensive)
2. Detailed calibration of the model against all field data (similar to «data assimilation» in meteorological models)
Use of Computer Modelling outlines of observationalComputer Modelling network | 12 Implementation | the chosen mass balance model
Approach by Huss et al. (2009, 2010), simplifying the basic processes and using the available data:
Melt: e.g. degree day model including potential solar radiation according to Hock (1999), calibrated against field data
Accumulation: precipitation record and accumulation measurements, calibration against field data Use of Computer Modelling outlines of observationalComputer Modelling network | 13 Implementation | computing and calibrating accumulation
Accumulation: combined use of precipitation record and accumulation measurements
퐶 푥, 푡 : accumulation
푃 푡 : precipitation
퐶 푥, 푡 = 푃 푡 ∙ 푐 ∙ 퐷 (푥) 푐푝푟푒푐: precipitation 푝푟푒푐 푠푛표푤 correction factor 퐷푠푛표푤(푥): spatial multiplier (normalized) 푥: location 푡: time
Calibration against measured snow accumulation (minimizing root mean square error) Use of Computer Modelling outlines of observationalComputer Modelling network | 14 Implementation | computing and calibrating melt
Melt: e.g. apply degree day model including potential solar radiation according to Hock (1999), numerical daily resolution
푀: melt
푓푀: melt factor 푟: radiation factors (푓 +푟 ∙ 퐼)푇, 푇 > 0°퐶 푀 = 푀 퐼: potential solar 0, 푇 ≤ 0°퐶 radiation 푇: air temperature
푑푇 air temperature : 푑푧 gradient
calibrated against measurements of annual mass balance (minimize Root Mean Square Error) Use of Computer Modelling outlines of observationalComputer Modelling network | 15 Results | example Findel glacier mass balance year 2011/2012
1. Winter (after calibrating cprec , 2. Annual (after calibrating fM , rsnow , rice, 푑푇 Dsnow(x) to minimize RMSE) to minimize RMSE) 푑푧
1 km croses denote measurements
source: M.Huss source: M.Huss Use of Computer Modelling Computer Modelling | 16 Results | Versatile model output
modelled temporal evolution of surface mass balance for each individual point
source: M.Huss
source: modified after M.Huss Use of Computer Modelling outlines of observationalComputer Modelling network | 17 Results | Versatile model output
Computed glacier wide mass balance all years
dashed lines denote dates for fixed date surveys
points denote measurements
source: L. Sold Modelled transient cumulative mass balance allows adjusting floating date measurements to fixed date or stratigraphic system Use of Computer Modelling outlines of observationalComputer Modelling network | 18 Larger scale modelling | example Machguth et al. (2009)
• Previous approach is Area of Swiss glaciers (year 2000) ~950 km2 data intensive • availability of gridded «meteorological» data (e.g. Regional Climate Model output, re-analysis) • Invitation to apply similar models for large scale modelling • Applicable in data scarce regions? Use of Computer Modelling outlines of observationalComputer Modelling network | 19a Larger scale modelling | use of gridded data
• Application of downscaling approaches provides daily high-resolution grids of meteorological parameters • Use of meteorological measurements to bias- correct meteorological fields • Bias-corrected daily mean annual P (1979-2003), Regional Climate
meteorological fields are Model (RCM) output, native resolution input to surface mass balance model Use of Computer Modelling outlines of observationalComputer Modelling network | 19b Larger scale modelling | use of gridded data
• Application of downscaling approaches provides daily high-resolution grids of meteorological parameters • Use of meteorological measurements to bias- correct meteorological fields • Bias-corrected daily mean annual P (1979-2003), Regional Climate
meteorological fields are Model (RCM) output, native resolution input to surface mass balance model mean annual P (1979-2003), downscaled to
100 m resolution Use of Computer Modelling outlines of observationalComputer Modelling network | 20a Larger scale modelling | model output
ELA ≈ 2600 m a.s.l.
ELA ≈ 3300 m a.s.l.
modeled mean annual mass balance distribution Oct. 1979 – Oct. 2003 Use of Computer Modelling outlines of observationalComputer Modelling network | 20b Larger scale modelling | model output
ELA ≈ 2600 m a.s.l.
ELA ≈ 3300 m a.s.l.
modeled mean annual mass balance distribution Oct. 1979 – Oct. 2003 Use of Computer Modelling outlines of observationalComputer Modelling network | 21 Larger scale modelling | model evaluation
pronounced bias (!) in accumulation
SMB measurements are not used for calibration, only for model evaluation Use of Computer Modelling outlines of observationalComputer Modelling network | 22a The essence
• Importance of field data to calibrate models Accumulation processes • Cumulative mass balance Findel glacier (2004-2014) (back to the Findel glacier example...) • Comprehensive field measurements are essential
Geodetic mass balance
source: L. Sold Use of Computer Modelling outlines of observationalComputer Modelling network | 22b The essence
• Importance of field data to calibrate models Accumulation processes • Cumulative mass balance Findel glacier (2004-2014) (back to the Findel glacier example...) • Comprehensive field measurements are essential
Geodeticpronounced mass balance offset in model output because of excluding accumulation data from model calibration
source: L. Sold Use of Computer Modelling Conclusions | 23
Conclusions Use of Computer Modelling outlines of observationalConclusions network | 24 Comparison | traditional approach vs. computer modelling
Traditional (manual contour line) Modelling
• Limited temporal and spatial • High temporal and spatial resolution of surface mass balance resolution of surface mass balance • Direct use of expert knowledge • Indirect use of expert knowledge • Dependent on availability and • Strongly dependent on availability quality of field data … and quality of field data • … but difficult to include all • Basal and internal mass balance available data can be modelled (uncertain) • Basal and internal mass balance • Applicable on large scales ... can only be estimated • … but how to calibrate/evaluate • Difficult to reproduce large scale applications? • Flexible with respect to inclusion of novel data/measurements
Comprehensive field measurements are essential. Thank You! Use of Computer Modelling outlines of observationalAppendix network | A Larger scale modelling Use of Computer Modelling outlines of observationalAppendix network | B Findel glacier measurements
source: L. Sold