Towards Understanding Pacific ITCZ Biases in Global Atmosphere Models

Tropical biases remain a significant problem in global atmosphere models, even at horizontal grid spacings of 5-­‐10 km. A well-­‐known example is the “double-­‐ITCZ” bias in the Pacific, which has been plaguing both coupled and uncoupled models for more than several decades now. Because this bias has been found even in global models that can be considered near “cloud-­‐resolving”, it would seem that other deficiencies than those involving convection schemes are also an important part of the problem. This idea is supported by our own current idealized simulations of the ITCZ, which are showing surprisingly strong sensitivity to the treatment of surface fluxes under light-­‐ wind conditions. In particular, schemes that predict weaker surface latent heat fluxes (LH) at low wind speeds tend to favor a double-­‐ITCZ pattern in rainfall, while a much more realistic pattern is favored using schemes that predict the opposite. Similarly, in real-­‐world reforecast simulations of the MJO, we are finding that model performance is substantially degraded as the amplitude of the “gustiness” effect on LH is made smaller. Here, we propose to extend these findings to a set of more focused reforecast simulations of the Pacific ITCZ using two different global models: 1) the NCEP Global Forecast System (GFS) and 2) the global Weather Research and Forecast model with a “super-­‐parameterization” for convection (SP-­‐WRF). The goal will be to characterize the lead-­‐time dependence of each model’s climatological ITCZ bias to changes in various aspects of the PBL/surface-­‐layer scheme, including the treatment of gustiness. Our hypothesis is that the amplitude of the double-­‐ITCZ bias will become smaller as the amplitude of the “gustiness” effect on LH is made larger, owing to enhanced availability of moisture on the equatorward flank of the ITCZ, where wind speeds are typically weaker. To allow direct statistical comparisons of the model output against NOAA R/V observations of LH, reforecasts will be generated for each of the past major field campaigns: EPIC, DYNAMO, TOGA-­‐COARE, KWAJEX, JASMINE, and Nauru99. Also, as a further test of our hypothesis, we will explore the statistical relationship between LH and wind speed in the set of global models that have contributed data to the CMIP5 archive. On the observational side of the problem, we will revisit the limited set of NOAA LH observations collected under weak wind speeds, to understand why direct (covariance) estimates are systematically smaller than indirect (inertial-­‐dissipation) estimates. Ultimately, the goal will be to develop an improved physically based treatment of gustiness effects in global models.