Strengths and Limitations of First-Generation Reanalyses for Understanding Climate Variability and Trends I
Presiding: R S Webb, NOAA Climate Diagnostics Center; J D Laver, NOAA Climate Prediction Center; R M Dole, NOAA Climate Diagnostics Center; P Arkin, Earth System Science Interdisciplinary Center, University of Maryland at College Park
A21B-01 08:30h
Reanalysis and the Climate Change Science Program
One of the key early deliverables of the U.S. Climate Change Science Program (CCSP) is a synthesis and assessment report on "Reanalyses of historical climate data for key atmospheric features; implications for attribution of causes of observed change". The significance of this deliverable is in the premise that "understanding the magnitude of past climate variations is key to increasing confidence in the understanding of how and why climate has changed and why it may change in the future". One component of the deliverable is a summary report on the current generation of re-analysis of historical climate data. The summary report will address such questions as: What aspects of climate trends do different re-analysis agree and disagree on? How do different re-analysis products compare to other independent assessments of climate variability and trends? What discontinuities and spurious trends and variability exist in the current generation of reanalysis products and for what reasons? How are global and regional climate trends and variability in surface temperature and precipitation related to trends in atmospheric circulation? What are the uncertainties in the current generation of re-analysis products and how can they be minimized. In this presentation, we review our current understanding of some of these issues. In particular, we will examine 1) the transition to the satellite era in the middle to late 1970s, 2) the agreement between reanalyses and independent observations in the trends of surface temperature and 3) the differences in the intensity and pattern of the hydrological cycle in re-analyses and independent estimates.
A21B-02 08:45h
Seasonal and Interannual Variations of Global Precipitation In the NCEP CDAS, CDAS2, GFS and CFS
Seasonal and interannual variations of global large-scale precipitation generated by the NCEP CDAS, CDAS2, GFS AMIP runs and CFS free runs have been examined and compared with those in the CMAP data set. The mean annual cycle of monthly precipitation is defined for a 23-year period from 1979 to 2001 for the CDAS, CDAS2, GFS AMIP runs and the CMAP, and for the first 32 years of free run periods for the CFS. In general, the mean annual cycles of CDAS, CDAS2, GFS and CFS are consistent with the observations, especially over the Pacific Ocean. Certain consistent errors, however, are observed. The ITCZ is too weak in CDAS and too strong in the GFS. Over the Atlantic, the ITCZ in the CDAS, CDAS2 and CFS displaced southward during December - May, while that in the GFS AMIP runs located in the same places as in the observations but presents excessive rainfall especially over the western ocean. Interannual variability of precipitation associated with the ENSO, NAO, AO and PNA is then investigated for the data periods. ENSO composite anomalies are fairly realistic, although the GFS appears too intense. NAO/AO composite anomalies from CDAS/CDAS2/GFS/CFS are quite good in the North Atlantic, while they are less realistic near and south of the equator. To further understand the model performance, oceanic boundary condition and atmospheric circulation fields associated with the seasonal and interannual variations of precipitation are examined. Detailed results will be reported at the meeting.
A21B-03 INVITED 09:00h
ERA-40 and Climate Variations
Reanalyses have brought an important multivariate component to the understanding of climate variations on the global scale. All observations are used to constrain short-range background forecasts in order to create an optimum analysis by combining the 4D- information from the forecast and observations over a long period of time taking into account their error characteristics. The background forecast is also used to identify and eliminate erroneous or conflicting observations and to give adjustments for the possible biases in them. The reanalysis view of climate variations in contrast to the traditional climate statistics is a simultaneous view of all variables from a single dynamically consistent source. In addition to the analyzed variables the model integration from analysis time to the next gives products from physical processes and therefore contributes to studies of hydrological cycle and general circulation. ERA-40 reanalysis products are a significant improvement on those from the first generation ERA-15 reanalysis. The improvements have been partly due to the development of the underlying operational forecasting system, to the identification of deficiencies in ERA-15 and also due to a more comprehensive usage of space based and conventional observations. ERA-40 now represents low frequency variability and trends of surface air temperature that is in good agreement with independent analysis of surface station data since late 1970s. At the same time the monthly mean middle to upper-tropospheric and lower stratospheric temperatures are in good agreement with the MSU record processed by Mears et al. and the QBO is in good agreement with independent rocketsonde data. However, the Brewer-Dobson circulation is too intense, the age of air in the stratosphere is too young and significant temperature biases are found in upper stratosphere. Precipitation over the tropical belt is also excessive during the satellite era from 1979 onwards. Improvements towards the next generation climate reanalysis will be made by improving the assimilating model, using a more refined assimilation method, and by improving, where possible, the data coverage and data usage. This will yield reanalyses that are further suited to establishing a more objective view of variations in climate.
A21B-04 09:15h
Using Observations to Assess the Increasing Trend in Storm Track Activity Found in NCEP/NCAR and ERA40 Reanalyses Data
Based on analyzing reanalysis data, many recent works have shown that Northern Hemisphere baroclinic wave and cyclone activity in winter have increased significantly over the second half of the twentieth century. Studies have found that the average number of deep cyclones have increased both in the Pacific and Atlantic, and eddy energy and heat transports have also increased. However, since the storm tracks peak over the oceans, very little data are assimilated from 1948 to the early 1970s, except at the surface. Since the mid-1970s, much more aircraft and satellite data are available to constrain the reanalyses. Hence there is a possibility that such changes in data coverage could introduce biases in the trends found in reanalysis data. In this study, storm track activity computed from reanalysis data are compared to similar quantities computed directly from observations to assess whether there are biases in the reanalysis trends. For eddy activity at the upper troposphere, variances computed based on radiosonde and aircraft observations are compared to those computed based on NCEP/NCAR and ERA40 reanalyses. The results suggest a possible high bias in the trend computed from the NCEP/NCAR reanalysis, with the weaker trend seen in the ERA40 more consistent with the trend determined from observations.The high bias in the NCEP/NCAR reanalysis trend appears to be worst over the Pacific. The storm track trend near the surface has also been examined. Near the surface, the trends computed from the NCEP/NCAR and ERA40 reanalyses are comparable. However, both trends appear to be higher than trends computed based on COADS ship observations. Again, the biases appear largest over the Pacific. Taken together, the results suggest a significant upward trend in storm track activity over the Atlantic during the second half of the twentieth century, but the trend over the Pacific is less robust. Results of this study suggest that a "data consistent" reanalysis may be needed to better quantify the upward trend in Northern Hemisphere storm track activity.
A21B-05 09:30h
Arctic Performance of Recent Atmospheric Reanalyses
Recent global reanalyses from the European Centre for Medium-range Weather Forecasts (ERA-40) and the National Centers for Environmental Prediction (NCEP-2) are compared and contrasted for the region poleward of 50N during the "modern" satellite era (1979-2002). The focus is on the directly observed atmospheric variables, namely temperature, geopotential height/pressure, moisture, and winds. The question is posed: how well do these reanalyses reproduce on the monthly time scale the observations that were assimilated from rawinsonde stations? A more detailed analysis is conducted for the North American Arctic by also including the North American Regional Reanalysis (NARR), recently completed by NCEP. Both surface and free atmospheric validation studies are conducted. Notably, the reanalysis predicted precipitation amounts are examined in relation to corrected gauge measurements of monthly precipitation totals. One motivating factor is to examine the added benefit of the precipitation assimilation performed by NARR for at least some parts of the North American Arctic.
A21B-06 09:45h
North American Regional Reanalysis: Evaluation Highlights and Early Usage
Following an about a 6-year development and production effort, the NCEP 25-year (1979-2003) North American Regional Reanalysis (NARR) has been completed. It is continued in near-real time as a Regional Climate Data Assimilation System, R-CDAS. Done for a domain including all of the North American continent and considerable parts of the adjacent oceans, with a 32 km/ 45 layers resolution, the NARR has resulted in a consistent high resolution climate dataset suitable for numerous applications. As aimed for, comparisons against the available NCEP/NCAR Global Reanalysis (GR) show that its accuracy is generally considerably higher than that of the GR. This is demonstrated by looking at rms fits to rawinsonde observations (raobs), as well as fits to surface data. Assimilation of analyzed precipitation resulted in precipitation fields that to a very high degree replicate the analyzed fields, a feature boding well when it comes to land-surface and hydrology applications. Several lessons that should be useful in planning future reanalyses are mentioned. Early usage of the NARR datasets as reported on at the recent NARR Users Workshop, San Diego, CA, is summarized. It includes additional validation work, uses for water and energy budget assessments in relation to predictability and climate variability issues, efforts at identification of useful severe weather predictors, North American monsoon studies, use for driving regional climate simulations, and more.