• Tidak ada hasil yang ditemukan

The purpose in the present thesis is to evaluate overall performances of precipitation, cloud and radiation processes of various convective parameterizations (CPs) in single column model (SCM) by comparing with explicit cloud-resolving model (CRM) and their interactions during the Southern Great Plain during 1997 (SGP97) on warm season. It is a necessary part of designing improvement of CPs how to determine the role of different concepts of CPs impact on the model performance. For the goal of the thesis, 9 CPs available in a single version of column in WRF and the explicit model, the Goddard Cumulus Ensemble (GCE) model as the reference CRM, are applied with the same microphysics (MP) and radiation physics (RAD) from CRM. In addition, they are also prescribed by the same SGP forcing data which includes the large-scale advection terms of dry static energy/moisture and surface fluxes.

All the runs simulate ensembles with 13 different initial conditions by interpolating 30 minutes of 3 hourly data from the first initial condition to the next step. It should be manifested that the feedbacks among precipitation, cloud, and radiation are controlled with the same CP, MP, and RAD by simulating SCM and CRM.

CRM represents good performances of precipitation patterns and skill scores as the reasonable reference model, driven by intensive observational forcing on continental summer season. And previous studies have also proved the performances (Xu et al. 2002; Lee et al. 2010). Compared to CRM, there are common features of CPs below,

1) CPs shows diversities of generating precipitation, cloud-radiation processes, and simple scores due to different type of trigger function and closure assumption.

2) In the simple score, the forecast skill (PSS; Peirce Skill Score in the present thesis) relies on not only hit rate (HR), but also false alarm rate (FAR). It means that it is important for triggering precipitation to be represented in large-scale model for grant, and non-triggering precipitation should be also considered for simulating precipitation. It is necessary for the understanding of natural precipitation.

3) Strong threshold of triggering CPs (e.g. KF and SAS-type) less simulates light precipitation, while the strong criteria of triggering helps them to simulate strong convection. Weak criteria of triggering CPs tends to make light rainfall. There is the balance of precipitation intensity between the closure and the trigger.

4) CPs estimate strong condensation and heating/cooling of shortwave (SW)/longwave (LW) radiation rate. This is because of larger amount (~20%) of cloud fractions implying that the strong diabatic heating in convections interplays cloud-radiation processes. Relative humidity is shown for moist state of CPs. It is also favorable state for representing large clouds covering

in a model column.

5) There are two types of CP simulations in this thesis; convection-type (KF, OSAS, NSAS, ZM, BMJ, GD, and GF) and grid-type (OKF and TIED) defined by how much convective precipitation is simulated. Precipitation amount is largely estimated in convection-typed runs more than grid-typed ones. On the other hand, hydrometeors of convection-typed simulations are predicted less than grid-typed ones. It is similar for release of latent heat to generate condensation rates in both convection- and grid-type, albeit the heating of SW is more sensitive.

In nature, grid-typed anvil clouds fall more precipitation down on the surface than deep convections. In one model column, because, the amount of precipitation from grid-scale clouds are quantitatively simulated smaller than convective systems, even though the model state is more humid. Note that there is unbalanced relation of grid-typed runs between precipitation amount and cloud-radiation feedbacks.

6) Grid-typed runs simulate larger amount of hydrometeor than other models. Commonly, all the CPs underestimate graupel because CPs simulate weak vertical updraft not to reach the strong criteria conversion from snow to graupel. This is the reason why all ice particles are underestimated; the weak upward transport of moisture. Almost CPs overly have cloud water particles and this is because of more humid state.

7) CRF at the top of atmosphere is also the same diversity and overestimated by LW and SW CRF component driven by larger cloud fractions; the magnitude of CRF is substantially different across CPs. Grid-typed runs are stronger CRF than convection-typed ones. This is because SCM conceptually assumes that anvil clouds contributing to grid-scale rainfall are larger coverages than convective systems such as in nature.

8) Diurnal cycle of precipitation also shows diversity of maximum patterns on different time due to their deficiencies. CPs well-simulates the diurnal cycle on the afternoon. But, most of CPs have a systematic bias in the timing of rainfall during a day, with more erroneous daytime rainfall. This is because surface fluxes get heating on the surface.

In this thesis, SAS-type CPs adopt reasonable assumptions for closure and the trigger. For the closure assumption, SAS-typed schemes generate stronger upward transports than other convtection-typed ones.

Because they show larger simulating ice particles (e.g. snowflakes) near cloud-base and relatively smaller cloud waters. But, NSAS is sensitive for the intensity of rainfall. In the view of the trigger function, SAS-typed physics well-simulates nocturnal peak because of the convective inhibition adopted in the trigger function.

The ensemble is one of promising ways to improve model performances to minimize uncertainties from the assumptions of CPs (Hawkins and Sutton 2009; Knutti et al. 2010). One of ensemble method, in this thesis, is multi-convective ensemble (MCE). Using WRF, MCE has widely used to improve forecast by diminishing uncertainties through model physics. For ensemble for multi-physics, CPs are

more sensitive that other physics (Jankov et al. 2005). And the present study shows MCE averaged by nine CPs. The results imply that MCE is reliable to reduce the uncertainties from CPs. In certain part, MCE is more believable result than CRM, compared to observation. For example, the case of diurnally- maximum precipitation on afternoon shows that MCE reproduces observational signal more properly than CRM. Consider that the ensemble is not perfect because multi-models share their own bias. There are examples in this experiments that interactions of CP, MP, and RAD are diverse in clouds, and deficiencies from CPs play a role to simulate different precipitation and vertical distributions.

Even though the simulations of CPs in cloud systems are the same set-up, deficiencies from the assumptions of CPs cause diverse spreads of precipitation and vertical profiles for cloud and radiation processes. Compared to the explicit CRM, interactions among precipitation, cloud, and radiation do not balance with each physics in limited convective systems. A possible step to improve CPs need to stabilize the relation of convections with other different schemes such as MP and RAD. Then, interactive simulations need to investigate the understanding of precipitation and cloud-radiation processes because SCM is non-interactive mode which is not able to analyze physical and dynamical response in the large-scale model systems. The solution to develop large-scale models will be closer to the explicit model state or the nature.

Dokumen terkait