PACS/ECMWF intercomparison

1. Introduction

There are two types of data: (i) the direct observations e.g. wind, temperature, humidity and radiation, and (ii) the derived parameters as skin temperature and turbulent fluxes. The algorithms used for the derived products are not the same as used in the ECMWF model, but the differences between algorithms is small and the effect of the algorithms tends to be much smaller than the model errors and the representativeness errors of the observations. The data is fairly complete and looks very reasonable. It is probably the best quality data one can expect from long term monitoring on buoys.

To compare the operational model with observations, 10 m wind, 2m T and q, SST and surface fluxes were retrieved for the two buoy locations. In this intercomparison 6, 12, 18 and 24 hour forecasts are used from the 12 UTC analysis. In accordance with the archiving convention, the fluxes are averaged over 6 hour intervals and the other parameters are instantaneous values. The observations are used consistently with the model: the fluxes are averaged over 6 hour interval and the other parameters are sampled every 6 hours (sampling here means the observations nearest to 0, 6, 12 or 18 UTC are used only). The basic PACS data consists of one hour averages.

A series of intercomparisons has been made: time series, scatter plots with stratification of points according to forecast time or precipitation. During the SSMI esuite both suites are compared and total column water vapor is included. Also the effects of the 21r3 physics changes are documented using a 3 week data assimilation experiment. The intercomparison has resulted in a large series of plots which are listed in the following together with a few conclusions:

2. Acknowledgement

3. Time series

• t2 (2m temperature) north buoy and south buoy
• q2 (2m specific humidity) north buoy and south buoy
• SST (sea surface temperature) north buoy and south buoy
• tsk (sea skin temperature) north buoy and south buoy
• wind (wind speed) north buoy and south buoy
• dt (temperature difference between sea surface and 2m level) north buoy and south buoy
• dq (specific humidity difference between saturation value at sea surface and 2m level) north buoy and south buoy
• SSHF (sensible heat flux) north buoy and south buoy
• SLHF (latent heat flux) north buoy and south buoy
• SSR (net solar radiation; averaged over 24 hours) north buoy and south buoy
• STR (net surface thermal radiation) north buoy and south buoy
• CC (total cloud cover; model only) north buoy and south buoy
• Precip (total precipitation; model only) north buoy and south buoy

1. See SST south buoy for the LaNina temperature drop in 1998. It results in a negative sea-air temperature difference and stable stratification due to warm air advection. Also note the spikes in the the observed temperature difference which the model does not have. These spikes have to do with a temperature drop due to precipitation events which can be due to evaporation of precip. However, it can also be an artefact of the observations (temperature observations during rain are difficult, because the instrument may become wet resulting in a wet bulb rather than a dry bulb).
2. The model underestimates solar radiation predominantly during periods with precipitation and overestimates latent heat flux . Wind speed is slightly underestimated.
3. The overestimation of latent heat flux is the direct effect of a too dry boundary layer i.e. the humidity difference between sea surface and atmosphere is too large in the model. Too dry model boundary layers were also seen during TOGA/COARE with ERA-15 whereas the operational model in 1993 was much better. This result suggests that the dryness was introduced with the new cloud scheme and the related convection changes.

4. Scatter plots (6 and 24 hour forecasts)

• wind (wind speed) north buoy and south buoy
• dt (temperature difference between sea surface and 2m level) north buoy and south buoy
• dq (specific humidity difference between saturation value at sea surface and 2m level) north buoy and south buoy
• SSHF (sensible heat flux) north buoy and south buoy
• SLHF (latent heat flux) north buoy and south buoy
• SSR (net solar radiation; averaged over 24 hours) north buoy and south buoy
• STR (net surface thermal radiation) north buoy and south buoy

1. There are no obvious indications of spinup in these diagrams.

5. Scatter plots (high and low model precipitation)

• wind (wind speed) north buoy and south buoy
• dt (temperature difference between sea surface and 2m level) north buoy and south buoy
• dq (specific humidity difference between saturation value at sea surface and 2m level) north buoy and south buoy
• SSHF (sensible heat flux) north buoy and south buoy
• SLHF (latent heat flux) north buoy and south buoy
• SSR (net solar radiation; averaged over 24 hours) north buoy and south buoy
• STR (net surface thermal radiation) north buoy and south buoy

1. Wind shows large scatter particularly at low wind speeds (meso-scale variability?).
2. Solar radiation is underestimated. Such an underestimation is understandable in cases that the model has precipitation, but it also happens during periods with no or little precipitation.
3. The temperature difference between sea surface and atmosphere shows too little variability in the model.

6. Scatter plots for esuite (E12) with SSMI (6 and 24 hours)

• wind (wind speed) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• dt (temperature difference between sea surface and 2m level) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• dq (specific humidity difference between saturation value at sea surface and 2m level) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12

1. Comparing the operational results without SSMI and the the esuite with SSMI shows that SSMI slightly increases boundary layer moisture. There is no obvious difference between 6 and 24 hour forecasts.

7. Scatter plots for esuite (E12) with SSMI (high and low precip)

• wind (wind speed) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• dt (temperature difference between sea surface and 2m level) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• dq (specific humidity difference between saturation value at sea surface and 2m level) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• tcwv (model total column water against qs-q2) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• tcwv (model total column water against q2) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12

1. Comparing the operational results without SSMI and the the esuite withSSMI shows that SSMI slightly increases boundary layer moisture. The link to precipitation is not obvious.

8. Scatter plots for L60+new_clouds experiment (zy5h; 12 and 24 hours)

• wind (wind speed) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• dt (temperature difference between sea surface and 2m level) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12
• dq (specific humidity difference between saturation value at sea surface and 2m level) north buoy OPR and north buoy E12 and south buoy OPR and south buoy E12

1. Comparing the sea-air humidity difference of the operational model and experiment zy5h (21R3; L60+clouds) shows the experimental version increases boundary layer moisture.
2. At the same time the sea-air temperature difference is also improved. Compare the operational model and experiment zy5h (21R3; L60+clouds) .
3. The main reason for the improvement is the increased evaporation of precipitation in the experimental model version, which cools and moistens the boundary layer.