Low (web) Quality - BALTEX
Low (web) Quality - BALTEX
Low (web) Quality - BALTEX
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
171<br />
Analysis of surface air temperatures over Ireland from re-analysis data and<br />
observational data<br />
Priscilla A. Mooney 1 , Frank J. Mulligan 2 and Rowan Fealy 1<br />
1 Irish Climate Analysis and Research Units, National University of Ireland Maynooth, Kildare, Ireland.<br />
priscilla.a.mooney@nuim.ie<br />
2 Department of Experimental Physics, National University of Ireland Maynooth, Kildare, Ireland.<br />
1. Abstract<br />
Since the observational data used in both ERA-40 and<br />
NCEP/NCAR Reanalysis data are largely identical, it is<br />
frequently assumed that either dataset is equally valid in the<br />
assessment of climate models. However, comparisons of<br />
these two dataset shows that subtle differences do exist and<br />
it is important to compare them to observational data in the<br />
region of interest to determine which reanalysis dataset<br />
should be used.<br />
In this study, surface air temperatures at 2 m altitude<br />
predicted by ERA-40 and NCEP/NCAR reanalysis (NNRP-<br />
1) have been compared with observations at eleven synoptic<br />
stations in Ireland over the period 1958-2000. Both<br />
reanalysis datasets show good agreement with the observed<br />
data and with each other. Slopes of the least-squares line to<br />
scatter plots of reanalysis data to observational data show<br />
small differences between the two reanalyses, with NNRP-1<br />
slopes and ERA-40 slopes ranging between (0.75-0.97)<br />
±0.01 and (0.79-1.0) ±0.01, respectively. Summary<br />
statistics and the monthly mean temperatures over the 1979-<br />
2000 period showed that both reanalyses predicted<br />
significantly warmer winters than the observations which<br />
caused the slopes of the best fit lines to be consistently less<br />
than unity.<br />
2. Introduction<br />
Over the past decade reanalysis data has found widespread<br />
application in many areas of research ranging from studies<br />
of climatic trends (Ciccarelli et al, 2008) and climate<br />
modeling (Fealy and Sweeney, 2007) to estimation of<br />
renewable energy resources (Henfridsson et al, 2007). It is<br />
advantageous to use reanalysis data in certain research areas<br />
where observational data is sparse or when knowledge of the<br />
state of the atmosphere on a uniform grid is required.<br />
Some of the most well known reanalysis data sets are<br />
NCEP/NCAR (NNRP-1), NCEP/DOE (NNRP-2), ERA-15,<br />
ERA-40 and ERA interim. NNRP-1 and ERA-40 are two of<br />
the most widely used reanalysis archives. Although they use<br />
similar observational data, previous studies have shown<br />
subtle differences between them (Escoffier and Provost,<br />
1998; Li et al., 2004; Gleiser et al., 2005; Ruti et al., 2008).<br />
Simmons et al., 2004 compared surface air temperature<br />
anomalies from CRUTEM2v with ERA-40 and NNRP-1.<br />
They found that ERA-40 showed better agreement with<br />
CRUTEM2v than NNRP-1 over the time periods 1958-2001<br />
and 1979-2001. In addition they reported closer agreement<br />
between ERA-40 and CRUTEM2v over the period 1979-<br />
2001 compared with the period 1958-1979. This was<br />
attributed to the greater observational coverage after 1979.<br />
3. Surface Temperature Datasets<br />
NNRP-1 reanalysis data was produced by the National<br />
Centers for Environmental Predication (NCEP) in<br />
collaboration with the National Center for Atmospheric<br />
Research (NCAR). The assimilation system used in the<br />
NCEP/NCAR reanalysis is described in detail in Kalnay et<br />
al., 1996. The NNRP-1 data used in this study was obtained<br />
from the NOAA/OAR/ESRL PSD, Boulder, Colorado,<br />
USA <strong>web</strong> site at http://www.cdc.noaa.gov/.<br />
ERA-40 data was produced by the European Center for<br />
Medium range Weather Forecasting (Uppala et al, 2005)<br />
in collaboration with several other institutions with an<br />
interest in climate analysis and weather forecasting. The<br />
ERA-40 data used in this study was obtained from the<br />
ECMWF data server (http://data.ecmwf.int/data/).<br />
The observed station data used in this study for the period<br />
1958-2000 was obtained from Met Éireann, the Irish<br />
national meteorological service, for 11 synoptic weather<br />
stations. The synoptic stations, which are geographically<br />
dispersed around Ireland (figure 1), represent mixture of<br />
both coastal and inland locations. This study did not<br />
perform any homogeneity analysis of the data. However,<br />
the data is from the synoptic network which is manned by<br />
experienced meteorological officers and the data is<br />
considered to be of high quality.<br />
Figure 1. Map of Ireland showing the location of the<br />
eleven meteorological stations and the marine buoys M1<br />
and M4. The map is overlaid with grids for NNRP (red ---<br />
line) and ERA-40 (blue… lines).<br />
4. Collocation Method<br />
Spatially, the parameter values in the reanalysis data sets<br />
represent the value of that parameter in a ‘grid box’<br />
centred on the geographic coordinates given in the dataset.<br />
This presents a difficulty when comparing the station data<br />
with the reanalysis data, since the station data represents a<br />
single site within a grid box. Additional complications are<br />
introduced by the fact that the grid sizes in ERA-40 and<br />
NNRP-1 are different, the grid centres in the two<br />
reanalysis data sets are not coincident and the location of a<br />
station in the grid box can vary from the centre of the grid<br />
box to its edge. This is illustrated in figure 1 where the<br />
grids used by ERA-40 and NNRP-1 are plotted together<br />
with the location of the stations and marine buoys.