US GLOBAL WATER AND ENERGY BUDGET STUDIES: A CONTRIBUTION TO CEOP

John Roads (SIO)

Link to NOAA Strategic Plan: NOAA’s Mission Goal 2: Understand Climate Variability and Change to Enhance Society’s Ability to Plan and Respond

RESEARCH OBJECTIVES AND SPECIFIC PLANS TO ACHIEVE THEM

The World Climate Research Program (WCRP) developed the Coordinated Enhanced Observing Period (CEOP). As part of CEOP, several global reference sites provided a number of in situ observations of water and energy budget study (WEBS) variables. Processed satellite data (geophysical variables) will also eventually be available at these sites. Model Output Location Time Series (MOLTS) Output from numerical weather prediction models is now available at these sites. In addition, international Numerical Weather Prediction and Research centers, including the Experimental Climate Prediction Center (ECPC) at Scripps, have also archived a more complete synoptic gridded output set and there may eventually be corresponding gridded satellite data for these sites. Developing the hydroclimatological output from these data sets has required a special effort. However, this community effort is now providing a wealth of data for analysis of the diurnal cycle. ECPC’s particular scientific goal was to understand what components of the global water and energy cycles could be accurately measured, simulated, and predicted at regional and global scales. In particular, we have now isolated strengths and weaknesses of our ECPC global and regional spectral models’ (G-RSM’s) description of the water and energy budgets, which has helped to further develop and improve ECPC’s G-RSM.

RESEARCH ACCOMPLISHMENTS

ECPC global model output data sets (gridded and MOLTS) that have now been provided to the international model output archive include: (1) the National Centers for Environmental Prediction / Dept. of Energy Reanalysis II (RII; L28T62 grid), and (2) ECPC’s GSM (L28T62) used in place of the RII model as part of an upgraded reanalysis (ECPC’s RII is actually an enhanced reanalysis of the original NCEP/DOE, which in turn is an upgraded version of the upgraded NCEP/NCAR RI). The SFM RII includes a number of improved parameterizations and is thus expected to provide a somewhat more realistic analysis than RII (or RI), although this still needs to be assessed. NCEP by contrast, which besides DAO and ECPC form the 3 major U.S. modeling contributions to CEOP, is providing data from their latest Global Forecast System (GFS) as well as from the original NCEP/NCAR reanalysis (RI). Again, similar output is being provided by a number of other international NWP centers. Besides the standard analysis variables available every 6 hours, 6-hour forecasts initialized from either an RII or SFM reanalysis are made every 6 hours, and once a day, at 1200 UTC, a 36-hour forecast is made, again with both the RII and SFM models. The forecast output is available every 3 hours. We have made a special effort to provide all of the CEOP/WESP requested variables and processes for the entire CEOP period (1 July 2001-31 December 2004). This output includes top of atmosphere, integrated and vertically varying atmospheric and surface water and energy-cycle processes and variables. It should be noted that gridded output is developed first and archived locally, and then MOLTS (41 CEOP sites) are extracted from the gridded data.

We used a two-part approach to analyzing the diurnal cycle in ECPC’s models, focusing first on the phase and then on the amplitude. Ruane and Roads (2006) presented an overview of the performance of the RII and the SFM reanalysis with respect to the phase of the diurnal cycle. Each component of the water and energy cycles (both at the surface and in the atmospheric column) was compared over North America between the two global reanalyses and the North American Regional Reanalysis (chosen as a high-resolution basis for comparison). In addition to establishing a diurnal climatology for these reanalyses, the study also identifies biases in the convective and land-surface parameterizations that lead to uniform phases across diverse regions. A budget approach also shows that the parameterization biases are transferred into the reservoir terms, leading (among other things) to large diurnal variations in precipitable water (Figure 2). Large differences are also evident in many of the budget components between land and sea points. Overall, the SFM improves on the RII in phase performance in this region.

To establish an analysis methodology for the reanalyses’ diurnal amplitude, we compared the frequency characteristics of three satellites’ precipitation products to our output over the 3.5-year CEOP period. The role of the diurnal cycle in different parts of the globe is analyzed using both Fourier and harmonic approaches. By dividing the spectrum of precipitation into three wide bands (covering periods shorter than 2 days, between 2 and 30 days, and longer than 30 days), the frequency contributions reveal that far more power lies in the high-frequency range of the spectrum than is explained by the harmonic fit of the diurnal and semidiurnal cycles (Figure 3). The RII does a much better job of reproducing observed precipitation frequency characteristics at all frequencies than does the SFM, particularly in the Tropics, where the SFM underestimates high-frequency power. We are preparing this paper for submission.

Regional Model Simulations

Transferability studies are a useful approach for evaluating the performance of a regional model under a broad range of different meteorological conditions. In collaboration with the CEOP/Water and Energy Simulation and Prediction, the Global Energy and Water-cycle Experiment (GEWEX) Hydrometeorology Panel Transferability Working Group (GHP/TWG) an international Inter-CSE Transferability Study (ICTS) was established to compare simulations of different regional models over seven regional domains associated with the eight GEWEX Continental Scale Experiments (CSEs). For these domains long-term simulations were carried out from July 1999 to December 2004. Up to now seven regional climate models are participating in ICTS. ECPC is participating in this international experiment with the ECPC RSM. Our long-term transferability runs for the seven domains have also been archived at the international CEOP model archive in Hamburg, Germany.

In addition to this multi-model—multi-domain intercomparison study within ICTS, Meinke et al. (2006) conducted several single-model—multi-domain studies to evaluate the ECPC RSM parameterization schemes under different meteorological conditions. Comparisons are carried out with CEOP reference site measurements and other global data sets, in particular International Satellitic Cloud Comparison Project (ISCCP), Global Precipitation Climatology Project (GPCP) and Hamburg Ocean Atmosphere Project / Global Precipitation Climatology Centre (HOAPS / GPCC) precipitation. During the last year a major focus was set on the evaluation of the ECPC RSM simulated precipitation. After estimating uncertainty ranges of both the model and the observations, model deficiencies in the amount of precipitation simulation have been identified for all model domains, except for the Mackenzie GEWEX Study (MAGS) region. Although the RSM simulates the seasonal evolution and the spatial distribution well, the RSM has an almost uniform positive bias (RSM greater than observations), except for the domain over the BALTic sea EXperiment (BALTEX) CSE, where the RSM has a negative bias. Most of the positive bias is either connected with the Intertropical Convergence Zone (ITCZ) convection or with monsoon convection (Southeast Asia). Stratiform precipitation is also excessive over high orography (see Figure 3). Summarizing the above, it can be stated that the over-prediction of precipitation is caused by convective as well as stratiform / dynamic precipitation. To avoid useless changes in the model and its parameterization it is important to first find the best available convection scheme for every single domain. This way the model is adjusted to the meteorological conditions of a particular domain without tuning.

Since the control simulations used the Relaxed Arakawa Schubert scheme (RAS), sensitivity tests with 3 additional convection schemes were then carried out to see if the simulations could be improved. The additional convection schemes included: 1) the Simplified Arakawa Schubert scheme (SAS), 2) the Kain-Fritsch scheme (KF), and 3) the National Centers for Atmospheric Research (NCAR) Community Climate Model (CCM) scheme. It was found that the precipitation simulation could be significantly improved for almost all domains using either the KF scheme or the SAS scheme (Figure 4). The best results for the ITCZ convective precipitation were achieved with the SAS convection scheme. For the monsoon convective precipitation the KF convection scheme had the best results.

Conclusions and Recommendations

We have achieved three technical and data developments that will be useful for further research: (1) the global NCEP/DOE Reanalysis II procedure has been rerun for the CEOP time period in order to save an expanded list of water and energy variables; (2) the RII model was replaced with the ECPC Seasonal Forecast Model and the entire reanalysis procedure was rerun for the period 1998-2005, which encompasses the reanalysis period; and (3) the ECPC Regional Spectral Model (RSM) was successfully run in climate mode for 7 continental scale regions, which encompassed most of the global land regions. Figure 1 shows the data now stored at MPI.

To provide best possible data for the Inter-CSE Transferability Study (ICTS), new long-term runs for the seven domains have been started with exchanged convection scheme. Since in ICTS fixed model setup is required for all domains, the same convection scheme has to be applied for all runs. For the new long-term runs the SAS convection scheme was chosen because the accumulated biases of all domains are smaller than if the Kain-Fritsch convection scheme was used. In the future long-term runs for ICTS are planned and for each particular domain the best-suited parameterization schemes can be chosen. Before, however, further evaluation activity of variables other than precipitation and more sensitivity tests are needed.

In addition to the global analyses and regional simulations, we are also planning to develop long-term simulations with the global model using different convective parameterizations. These simulations will be useful for comparison to the regional simulations and will be analyzed like the current global analyses, that is, with a focus on high frequency and diurnal variations. In addition to developing a deeper insight into diurnal processes, we are also hoping to further improve the ECPC global to regional modeling system.

Fig. 1 Model data submitted to MPI archives. Note that we submitted 2 global model and 1 regional model set of data for entire CEOP phase 1 time period.

Figure 2: Percentage variance of July 2001 – December 2004 SFM reanalysis precipitation described by (a) low (period > 30 days) (b) Synoptic (period between 2 – 30 days) and (c) high (period < 2 days) frequencies, as well as (d) the diurnal and semidiurnal harmonic reconstruction.

Fig. 3 Monthly precipitation sum; Left: RSM, right: GPCP

Fig. 4 Differences in precipitation between RSM with different convection schemes minus GPCP, area means during the test months