A study of the shortwave schemes in the Weather Research and Forecasting model

Abstract

The radiative transfer cannot be explicitly resolved in the atmospheric models for two reasons: i) a full treatment of the radiative transfer equation (RTE) requires a high amount of computational resources and ii) the radiative transfer fields such as the optical thickness are not a direct solution of the Euler equations and hence, they must be parameterized as a function of the meteorological fields. Consequently, the physical processes related with radiation are simplified and approximated in physical schemes.

In the particular case of the solar radiation, the use of these parameterizations were reduced for many years to represent the day/night cycle inside the model. Therefore, the accuracy of the solar schemes was left in the background and the computational resources were prioritized. With the growth of the solar energy industry during the last decade, a paradigm shift has occurred. Now, the solar irradiance (i.e. global horizontal GHI, direct horizontal DHI and diffuse DIF) becomes an important product for resource assessment as well as for forecasting applications.

The main objective of this thesis is the identification and quantification of the sources of error that have a direct or an indirect contribution to the accuracy of the solar schemes, particularly, in those available in the Weather Research and Forecasting (WRF-ARW) model, widely used in the sector.

First, the thesis presents a review of the set of physical approximations considered in six solar parameterizations available in the WRF-ARW model: Dudhia, Goddard, New Goddard, Rapid Radiative Transfer Model for General Circulation Models (RRTMG), Climate Atmospheric Model (CAM) and Fu-Liou-Gu (FLG).

The sources of error are limitations in the representation of the radiative transfer as a conse- quence of the set of approximations assumed by one scheme. In this thesis three sources of error are analyzed: i) errors due to the vertical discretization of the atmosphere in a set of layers that are assumed to be homogeneous (truncation error), ii) errors due to the misrepresentation of the layer between the top of the model (TOM) and the top of the atmosphere (TOA), called TOM error and iii) errors due to the physical simplifications and parameterizations in the RTE, named physical error.

In order to avoid the uncertainty introduced by the other components of the model, the source code of each one of the six solar schemes has been separated of the model and adapted for working with 1-dimensional vertical profiles.

The studies of the truncation and TOM errors are performed by using ideal vertical profiles under four scenarios: a dry atmosphere, a wet cloudless sky, low water cloud and a high ice cloud. The results for the ETOM show that for the typical range of TOM values in mesoscale appli- cations (i.e. 10 hPa), the error with respect to a full atmospheric column is less than 0.5% and

hence, the TOM error can be neglected.

The analysis of the Etrun reveals that the sensitivity of the solar schemes on the vertical config- uration (i.e. number of vertical levels and their distribution) is directly related with the method used for the vertical integration of the multiscattering processes. For the typical mesoscale config- urations, the Etrun under clear-sky conditions is determined around 1.1%, 0.9% and 4.9% for the GHI, DHI and DIF, respectively. In both cloudy scenarios, the Etrun increases significantly, being more important for the high clouds.

The Ephys is analyzed under clear-sky conditions using real soundings from the Integrated Global Radiosonde Archive data-set and comparing the irradiance outcomes with the Baseline Solar Radiation Network measurements. With the exception of Dudhia, the behavior for all the parameterizations is the same. A large overestimation of the DHI with a large underestimation of the DIF that leads to a near-zero bias for the GHI. Polar sites show the lowest errors with a mean MAE of 2.1%, 5.2% and 3.7% for GHI, DHI and DIF, respectively. Midlatitude sites show the worst results with a mean MAE of 3.4% in GHI, 11.6% in DHI and 7.8% in the DIF.

L’objectiu principal d’aquesta tesi ´es la identificaci´o i quantificaci´o de les fonts d’error que tenen una contribuci´o directa o indirecta en la precisi´o dels esquemes solars, particularment en aquells disponibles en el model Weather Research and Forecasting (WRF-ARW), `ampliament emprat en el sector de l’energia solar.

Les fonts d’error s´on limitacions en la representaci´o del transport radiatiu com a consequ¨`encia del conjunt d’aproximacions assumides per cada esquema. En aquesta tesi hi ha tres fonts d’error que s´on analitzades: i) l’error degut a la discretitzaci´o vertical de l’atmosfera en un conjunt d’estrats que s’assumeixen homogenis (error de truncament, Etrun), ii) l’error com a resultat d’una repre- sentaci´o insuficient de l’estrat entre el cim del model (TOM) i el cim de l’atmosfera (TOA), anomenat error de TOM Etom, i iii) l’error degut a les simplificacions i a les parametritzacions f´ısiques de l’RTE, definit com a error físic, Ephys.

Per tal d’evitar la incertesa introdu¨ıda pels altres components del model, el codi font de cadas- cun dels sis esquemes solars ha estat separat del model i adaptat per treballar amb perfils verticals 1-dimensionals. Mitjan¸cant aquest m`etode, les habilitats dels esquemes solars poden ´esser anal- itzades sota condicions d’entrada id`entiques.

D’una banda l’error de TOM i el de truncament s’analitzen a partir de perfils ideals. De l’altra, l’error f´ısic s’evalua prenent dades de radiosondatge com a perfil vertical i comparant les sortides dels esquemes radiatius amb mesures en superf´ıcie.

Els resultats d’aquesta tesi mostren que l’Etom esdev´e negligible per la majoria d’aplicacions de mesoscala. Per configuracions t´ıpiques del model, l’Etrun en condicions de cel ser`e es troba al voltant de l’1.1%, el 0.9% i el 4.9% per la GHI, DHI i DIF, respectivament. En el cas amb nu´vols augmenta de forma significativa. L’estudi de l’Ephys mostra una relaci´o significativa amb el contingut de vapor d’aigua i els aerosols.