CAPT simulations are analyzed along a cross-section in the Pacific Ocean, from California to the Equator. The cross-section over the Pacific Ocean encompasses several fundamental cloud regimes such as stratocumulus, shallow cumulus and deep cumulus, as well as the transitions between them. A climatology of cloud types from ISCCP data is shown in Figure 1.
Resolution: T42
Model timestep: 20 minutes
Vertical levels: 26 hybrid levels
Initialization
The forecasts are initialized every day at 00 UT. The initialization data sets are:
- ECMWF ERA40 reanalysis for 1998
- ECMWF operational analysis for 2003
Output
- Output every 3 hours
- 5-day forecasts
- Time period: JJA 1998 and 2003
We examine the mean forecast errors averaged over JJA defined as:
CAM3 (JJA) - ERA 40 (JJA).
Figure 2 shows the error in the mean forecast along the cross section from day 1 to 5 as well the mean climate error.
Figure 3 shows the 5-day evolution of temperature, moisture, and water condensate at selected locations. The change in T, q and qc is the instantaneous value at day1-5 minus its value at the initial time. The locations have been chosen to cover different cloud regimes (latitude 32 is the stratocumulus region, latitude 10 is the ITCZ, the latitudes 2 and 20 are two cumulus regions) and to cover various error behaviors. Notice that when averaging over a long period (such as JJA), the forecast errors in T and q are equivalent to the change in T and q.
Near the equator, the CAM3 forecasts become very warm compared to the ERA40 in the upper troposphere (errors up to 3K) and too cold and too dry in the lower troposphere. In this region, the CAM forecast T error forms rapidly within the first 24 hours with smaller changes in the following days. As we move on the cross section towards the California, the forecast error grows more slowly and its vertical structure changes drastically: near the latitude 20N, the model is too warm in the entire troposphere and the forecast error forms very gradually. In the stratocumulus region (latitude=32N), the lower troposphere gets very warm and dry near the level 900 mb.
The mean climate error defined as CAM3 (JJA) - ERA 40 (JJA) ressembles the mean forecast errors after 5 days, showing that most of the error forms rapidly within the first 5 days.
We look at the budget equations of the temperature and moisture to identify which terms in the equations are responsible for the forecast errors. We are using the notation described in Table 1. In this section, we show plots of the terms of the temperature and moisture equations averaged over the first 24 hours at the four locations described above.
Table 1: Budget equation of T and q. Separation of the equation in its different terms.
a. Stratocumulus regions (latitude 32N)
The stratocumulus region shows a strong warming and drying of the atmosphere around 900 mb. After 3 days of forecast, the T error is about 5K and the model has lost about 2g/kg of moisture. The budget equations shows that this error can be related to the dynamics. This does not necessarily imply that the dynamics itself is incorrect, this can be due to the fact that the other paramerizations are too weak.
A timeseries of the error shows that this strong warming/drying is not a consistent feature during the JJA. Instead we have alternating periods with and without strong warming/drying as shown in Figure 6. The terms in the moisture and temperature prediction equations are very different between these periods.
For example, we compare composites of the week Jul 22-28 (without strong warming) to the week Aug 1- 7 (strong warming).
T budget at day 1: Figure 7
T budget at level 24: Figure 8
Horizontal advection of T at level 24: Figure 9
T budget at day 1: Figure 10
T budget at level 24: Figure 11
Horizontal advection of T at level 24: Figure 12
b. ITCZ (latitude 8N)
After 1 day, the upper troposphere is too warm and the lower troposphere is too cold and dry. In this region, the forecast error forms rapidly which is a signature of the deep convection. The budget plots show that the deep convection parameterization is driving the model to the wrong state. This does not necessarily imply that the convection paraneterization is incorrect, but it could be responding to incorrect forcing.
c. Latitude 2N
After 1 day, the upper troposphere is too warm and the lower troposphere is too cold and dry. However, the terms that are causing the warming of the upper troposphere are different during the JJA period. In June, the deep convection is responsible of the warming while in August it is the dynamics that is responsible of the warming.
d. Latitude 20N
The forecast is too warm and too dry above 900 mb. The forecast errors form very gradually and are still growing after 5 days of forecast. The shallow convection is active and it dominates the paramaterization term under 900 mb.
T budget: Figure 19
Q budget: Figure 20
More plots can be find at: http://www.cgd.ucar.edu/ccr/CPT/GPCI/GPCI_plots.html
