daily maps of minimum and maximum temperature for … · microsoft powerpoint - presentasjon1...
TRANSCRIPT
NEW METHOD:
• Use 1000hPa geostrophic wind speed as additional predictor (cube root works well)
(3)
(ugeo = 1000hPa geostrophic wind speed at station from NCEP Reanalysis)
• v1, v2, v3, v4, v5, v6 and const estimated for each month of year separately Set Tde-trended(station) = 0 in Eqn. (3)
• Compute hypothetic station temperatures with Tde-trended(station) = 0 in Eqn. (3) (denoted Tde-trended infigures below) to check quality of fit
Daily temperatures (only January shown)
2. Parameter estimation:
• v1, v2, v3, v4, v5 and const estimated for eachmonth of year separately Set Tde-trended(station) = 0 in Eqn. (1)
• Use normal monthly temperatures, monthlytemperatures (not shown) or daily temperatures
• Compute hypothetic station temperatures withTde-trended(station) = 0 in Eqn. (1) (denoted Tde-trended infigures below) to check quality of fit
a) Normal Temperatures (only January shown)
b) Daily temperatures (only January shown)
DAILY MAPS OF MINIMUM AND MAXIMUM TEMPERATURE FOR NORWAY
Norwegian Meteorological Institute
Contact:
Matthias Mohr
Norwegian Meteorological InstituteMeteorology and Climate DivisionP.O. Box 43 Blindern, N-0313 Oslo
e-mail: [email protected]: +47 2296 3000Direct phone: +47 2296 3381
ORIGINAL METHOD:
1. Used for daily mean, min and max temperatures:
De-trending: (at each station)
(1)
(Tde-trended = de-trended station temperature, Tstation = observed station temperature, zstation = station altitude, zmean, 40 km = mean altitude within circle of 40km Ø surrounding station, zmin, 40 km = lowest altitude within same circle, lat/lonstation = latitude/longitude of station, const = constant)
Kriging: (interpolation from stations to grid points)
Two-dimensional ordinary kriging, monthly uni-directional variograms from normal temperatures (Tveito et al., 2000)
Tde-trended (station) Tde-trended(grid)
Re-trending: (at each grid point)
(Tde-trended(grid) = de-trended station temperatures interpolated to grid, zDEM,Norway = digital elevation model of Norway, zmean, 40 km = mean altitude within circle of 40km Ø surrounding grid point, zmin, 40 km = lowest altitude within same circle, lat/lonstation = latitude/longitude of grid point, const = constant)
Trend variables: (for whole of Norway)
constlonvlatvzv
zvzvTstationT
stationstationkm
kmmeanstationstationtrendedde
5440min,3
40,21)(
constlonvlatvzv
zvzvgridTT
cellgridcellgridkm
kmmeanNorwayDEMtrendeddemapfinal
5440min,3
40,2,1)(
CONCLUSIONS:
• New method performs on par, but not better thanoriginal method
• Max. temperatures predicted almost as good as daily means
• Min. temperatures predicted considerably worse
• Temperatures predicted best in spring/summer and worst in winter (temperature inversions)
• UTM coordinates seem to give better results for Norway than Lat/Lon (not shown)
More work is needed…
constuvlonvlatvzv
zvzvTstationT
geostationstationkm
kmmeanstationstationtrendedde
35440min,3
40,21
6
)(
Daily mean
temperature
Minimum temperature
Maximum temperature
Correlation 0.943 0.930 0.937
RMSE (°C) 2.202 2.780 2.083
Daily mean
temperature
Minimum temperature
Maximum temperature
Correlation 0.945 0.929 0.942
RMSE (°C) 2.161 2.808 1.994
OBJECTIVE:
• Daily maps of mean, minimum and maximumtemperature with 1km resolution for Norway
• Use all available temperature observations (140 -240 stations depending on year)
• Use multiple linear regression & residual kriging
• Use information on local terrain
• Use information on large-scale atmosphericcirculation (e.g. geostrophic wind speed)
Daily mean
Temperature
Minimum temperature
Maximum temperature
Correlation 0.945 0.929 0.942
RMSE (°C) 2.162 2.809 1.997
BACKGROUND:
• Week correlation between 1000hPa geostrophicwind speed and daily temperatures
• Correlation: R ≈ 0.5 in January and R ≈ -0.3 in July
Temperatures increase with increasinggeostrophic wind speed during winter
Temperatures decrease with increasinggeostrophic wind speed during summer
WE USE THIS FACTOR AS ADDITIONAL PREDICTOR
References:Mohr, M., 2009: Comparison of Versions 1.1 and 1.0 of Gridded Temperature and Precipitation Data for Norway, met.no note # 19/2009, see http://met.no/
Mohr, M., 2008: New Routines for Gridding of Temperature and Precipitation Observations for “seNorge.no”, met.no note # 08/2008, see http://met.no/Forskning/Publikasjoner/Publikasjoner_2009/
Tveito, O. E., E. J. Førland, R. Heino, I. Hanssen-Bauer, H. Alexandersson, B. Dahlström, A. Drebs, C. Kern-Hansen, T. Jónsson, E. Vaarby Laursen and Y. Westman, 2000 : Nordic temperature maps (Nordklim), DNMI KLIMA Report09/00, see http://www.smhi.se/hfa_coord/nordklim/old/rapport0900.pdf
EXAMPLE: The Coldest (?) Day in Norway
RESULTS:
• ”Leave-one-out” cross validation (using de-trending, kriging and re-trending (see left))
• 10 years of data (Jan 2001 – Dec 2010)
• Correlation and RMSE as quality measure (only January shown in tables below)
• ca. 140 - 240 stations per day (depending on year)
1.Original method
a) Use parameters from normal temperatures
b) Use parameters from daily temperatures
2. Use new method (including geostrophic wind speeds) for min and max temperatures