PERFORMANCE OF DIFFERENT PARAMETERIZATION

CONFIGURATIONS OF WRF-ARW MODEL DURING

HEAVY RAINFALL OVER MI OYA BASIN

Research Proposal

By
E/15/107 GAMAGE K.H.
E/15/376 WANASINGHE W.M.A.Y

Supervised by
Dr N.G.P.B. NELUWALA

Project: C/15/58

Department of Civil Engineering

Faculty of Engineering
University of Peradeniya
Peradeniya 20400
Sri Lanka
November 2020
 ABSTRACT

Short Term high-intensity rainfall is one of the hallmark signs of climate change.
Due to extreme rainfall, the Mi Oya river basin is heavily affected by seasonal flooding
annually. By constructing a numerical weather forecasting model for the entire basin
incorporated with proper reservoir operation, flood damage can be reduced to a considerable
amount. Numerical weather research forecasting has been carried out for several locations
of Sri Lanka, but still not sufficient to identify the most suitable physics schemes for the Mi
Oya river basin. The main objective of this study is to find an optimum physics scheme for
the Mi Oya basin, by comparing the forecasting data from the WRF ARW model and
observed Rainfall data in this study, input data for the WRF model will be obtained from the
NCEP-NCAR database and hope to validate the model output data with observed point
rainfall data of four gauging stations in the basin. In fine model tuning, hope to investigate
the impacts of microphysics schemes, cumulus schemes, land surface schemes, boundary
layer schemes, and long/shortwave schemes. Ultimately, by analyzing the results of this
study, hope to identify the best physical schemes to forecast heavy rainfall events in the Mi
Oya river basin.

ii
 ACKNOWLEDGEMENTS

It is a great pleasure to take this opportunity to thank all, who helped us to complete our
project under CE405, Civil Engineering Project-1 successfully. We like to give our sincere
gratitude to the Department of Civil Engineering, Faculty of Engineering, University of
Peradeniya, for affording such a great opportunity to us. First, we like to give our special
thanks to our supervisor Dr. N.G.P.B Neluwala who helped us in checking the entire progress
and helping to solve the problem during the research. We would like to special thank Prof.
J.J. Wijetunge and other panel members for the valuable comments. At last, to all of those
named above and to any others who may have been inadvertently omitted are extremely
thankful, to spend their valuable time during our project period.

iii
 TABLE OF CONTENTS

ABSTRACT ..................................................................................................................... ii

ACKNOWLEDGEMENTS ............................................................................................. iii

TABLE OF CONTENTS .................................................................................................. iv

LIST OF FIGURES .......................................................................................................... vi

LIST OF TABLES ........................................................................................................ viiii

LIST OF ABBREVIATION……………………………………………………………...viii

CHAPTER ONE ................................................................................................................ 1

INTRODUCTION .........................................................................................................1

1.1 Background ....................................................................................................1

1.2 Aim ................................................................................................................3

1.3 Objectives .......................................................................................................3

1.4 Scope ..............................................................................................................3

CHAPTER TWO ............................................................................................................... 4

LITERATURE REVIEW ...............................................................................................4

2.1 Introduction ....................................................................................................4

2.2 General Circuation model (GCMs) .................................................................5

2.3 Climate Downscaling ......................................................................................5

2.3.1 Statistical Downscaling………………………………………….….....6

2.3.2 Dynamically Downscaling…………………………………………….6

2.4 Dynamically Downscaling Models .................................................................6

2.4.1 REG CM Model…………………………………………………….....6

2.4.2 WRF Model………………………………………………………...…6

2.5 Physics Scheme……..……………………………………………………....….7

iv
 2.5.1 Microphysics scheme………………………………………….……....8

2.5.2 Cumulus scheme……………………………………………………....9

2.5.3 Land surface scheme…………………………………………….…...10

2.5.4 Longwave and shortwave radiation scheme…………………………10

2.5.5 Planetary boundary scheme………………………………………….11

2.6 Status of WRF Model in Sri Lanka …………………..……………………11

2.7 Input data…………………………………………………………………..15

CHAPTER THREE ......................................................................................................... 16

MATERIALS AND METHODS.................................................................................. 16

3.1 Study area .................................................................................................... 16

3.2 Data collection………………………………………………………………..17

3.3 Methodolgy brief……………………………………………………………..18

3.4 Model configuration………………………………………………………….19

3.5 WRF Pre-processing system………………………………………....…....….20

3.5.1 Geogrid………………………………………………………………20

3.5.2 Ungrid………………………………………………………………..20

3.5.3 Metgrid……………………………………………………………....21

3.6 Validation methods…………………………………………………………...21

CHAPTER FOUR…………………………………………………………………………23

4.1 Preliminary studies…………………………………………………………….23

Work schedule ................................................................................................................. 24

REFERENCES ................................................................................................................ 26

v
 LIST OF FIGURES

Figure 2.1 WRF ARW modelling system flow chart…...………...……………………….…7

Figure 2.2 Microphysics Parameterization Schemes ……………………………………….9
Figure 2.3 Observed 24-h rainfall on 16 May 2016 and Model output…………………….12
Figure 2.4 WRF experiments and Observed rainfall data on 20-12-2014………………….13
Figure 2.5 WRF predictions and observed rainfall with different Land surface model…….14
Figure 3.1 Methodology flow chart………………….…….……………………………….16
Figure 3.2 Mi Oya river basin……....…………………………………………………. ….17
Figure 3.3 Rainfall distribution over 24 hours on 06-11-2020………………………….…17
Figure 3.4 Selected WRF domain configuration in Weerakoon et al., (2016) ……….…….19
Figure 3.5 WPS Flow diagram…………………………………………………………….20
Figure 4.1 MobaXterm SSH terminal emulator…………………………………………....23

vi
 LIST OF TABLES

Table 2.1 Physical equation of NWP……………………………………………….……….4

Table 2.2 Strength and weakness of statical vs dynamically downscale…………………….5
Table 2.3 Summary of physical schemes which were used in research papers………...…...8
Table 2.4 Available GRIB data from NCAR……………………………………………….15
Table 3.1 Chosen physics schemes for the study…………………………………………...18
Table 3.2 Selected physics combinations………………………………………………….19
Table 3.3 Summary of evaluation equations………………………………………………22

vii
 LIST OF ABBREVIATIONS

NWP-Numerical Weather Prediction

WRF- Weather Research Forecasting
NEP-North East Monsoon
SWM-South West Monsoon
GCM-General Circulation Model
RCM-Regional Climate Model
NCEP- National Center for Environmental Prediction
NCAR-National Center for Atmospheric Research

viii
 CHAPTER ONE

INTRODUCTION

1.1 Background
‘It is far better to foresee even without certainty than not to foresee at all’ (Henri Poincare,
French mathematician). Weather forecasting has been improved during the last few decades
to foresee upcoming weather events so that farmers can know when to plant or harvest their
crops, people can choose where and when to take their holidays to take advantages of good
weather, regions can be evacuated if hurricanes or floods are expected and aircraft and
shipping rely heavily on accurate weather forecasting.

Weather forecasting is the prediction of the weather through the application of the principles
of physics, supplemented by a variety of statistical and empirical techniques. In addition to
predictions of atmospheric phenomena themselves, weather forecasting includes predictions
of changes on Earth’s surface caused by atmospheric conditions. E.g., snow and ice cover,
storm tides, and floods. (Cahir, 2019)
Numerical weather prediction (NWP) is a method of weather forecasting that employs a set
of equations that describe the flow of fluids. These equations are converted into computer
code and use governing equations, numerical methods, parameterizations of other physical
processes and combined with initial and boundary conditions before being run over a domain
(geographic area) (Weather.gov, 2020). These models are widely used nowadays in weather
forecasting services. The models that are commonly considered include the global models
and the mesoscale models with horizontal resolutions in the order of several kilometers to a
couple of tens of kilometers. The performance of NWP models with even higher spatial
resolutions is studied extensively recently with the objective of making a location-specific
forecast.
The two monsoon seasons are the southwest monsoon (SWM - summer monsoon) from May
to September and the northeast monsoon (NEM - winter monsoon) from December to
February. During the SWM season, heavy rainfall is received by the low land areas in the
south-western part of Sri Lanka and the western slopes of the hill country. The NEM brings
rainfall to the northern and eastern parts of the country. The two inter monsoon seasons are
from March to April (FIM-first inter monsoon season) and from October to November
(SWM - second inter monsoon season). Weather systems such as convergence, convection,
and depressions bring intense rain during the inter-monsoon seasons.
Many parts of Sri Lanka are vulnerable to extreme rainfall events such as floods and droughts
during the monsoon seasons (SWM and NEM) and occasionally due to tropical depressions
in the Bay of Bengal, particularly during the SWM season and the early part of the NEM
season.
The worst of the heavy rainfall struck from 09 September 2020, with flooding reported in
the districts of Puttalam and Gampaha, with strong winds and urban flooding affecting areas
around the capital, Colombo. Almost 30,000 people were affected by floods in Gampaha,
with over 17,000 of them in Wattala and 8,000 in Kelaniya. A total of 458 people was
displaced and staying in relief accommodation in the district. Meanwhile, around 2,000
people were affected by severe weather in Puttalam and 7,000 in Colombo District. (Flood
List News in Asia, News, 2020)

1
Determining the best Physics scheme for numerical weather prediction (NWP) is of great
importance and has been the subject of many recent studies by researchers from different
backgrounds. Not surprisingly these studies result in apparently rather different methods;
however, on closer inspection, they can usually be shown to be closely related. In recent
studies, several numerical methods are used to predict the weather in a selected area.
For example, Rodrigo et al., (2018) assessed Sensitivity Study of WRF Numerical Modeling
for Forecasting Heavy Rainfall in Sri Lanka, Weerakoon, et al., (2016) examined the
comparative study of WRF & REGCM weather predictions for the upper Mahaweli river
basin, Darshika, et al., (2017) simulated Heavy Rainfall During 19th to 28th December 2014
Using WRF for Different Atmospheric Physics and Nandalal et al., (2012) researched the
application of WRF weather model to forecast precipitation in Nilwala river basin.
According to those studies which were carried out in Sri Lanka and other studies overseas,
most researchers used the Weather research forecasting (WRF) model. By using this model,
they have gained comparatively good results than other numerical forecasting models. But
there is still no exact conclusion about physics schemes from the studies in Sri Lanka and
need further investigation and hope to fill that research gap by this study to a reasonable
extent.
In our study, we are going to predict the weather due to heavy rainfall by using Weather
Research & Forecasting models. Weather Research Forecasting (WRF) is the common use
of the numerical model for the research community. In this model Microphysics and
cumulus, parameters are the important parts of capturing the heavy rainfall.
The selected study area, the Mi Oya river basin is a 108 km long river, in northwestern Sri
Lanka. It begins in Saliyagama and flows northwest, emptying into the Indian ocean through
Puttalam. Mi Oya river basin has a 1516 km catchment area and annual precipitation volume
2

and discharge volume to sea as 2176×106 m3 and 338×106 m3 (Rajatane, 2003). Due to the
extreme weather conditions, the Mi Oya river Basin is heavily affected by seasonal flooding
annually. The spilling of Tabbowa and Inginimitiya reservoirs produces a major impact on
the flood damage (DMC Sri Lanka, 2019). By constructing a rainfall forecasting model for
the entire river basin incorporated with proper reservoir operation, flood damage can be
reduced to considerable amounts. In our study, we are hoping to identify suitable physics
schemes for two heavy rainfall events that occurred in the Mi Oya basin. (2016/05/13-
2016/05/16 and 2019/12/18-2019/12/20).
The outcomes of this research are useful for identifying the heavy rainfall event and could
be helpful to control the flood damage in vulnerable areas.

2
1.2 Aim

Forecasting heavy rainfall events in the Mi Oya basin

1.3 Objectives

• Set up a Numerical weather forecasting model

• Find optimum physics scheme for Mi Oya basin

1.4 Scope

• Weather Research Forecasting (WRF) model will be used as the Numerical weather
forecasting model
• WRF model result will be evaluated for two heavy rainfall events data (2016/05/13-
2016/05/16 and 2019/12/18-2019/12/20)
• Physics combinations from multiple Microphysics, Cumulus schemes with other
default schemes for PBL scheme, long wave /short wave radiation schemes, and
Land-surface scheme will be tested.

3
 CHAPTER TWO

LITERATURE REVIEW

2.1 Introduction

Weather is a mix of events that happen each day in our atmosphere. Weather is unalike in
different parts of the world and changes over minutes, hours, days, and weeks. Most weather
happens in the troposphere, the part of Earth’s atmosphere that is closest to the
ground (Scied.ucar.edu, 2020). climate describes what the weather is like over a long period
in a specific area. Different regions can have different climates. To describe the climate of a
place, we might say what the temperatures are like during different seasons, how windy it
usually is, or how much rain or snow typically falls.
There are a lot of weather forecasting methods in use, such as the analog method,
climatology method, statistical forecasting, persistence forecasting, numerical weather
prediction (NWP), etc. NWP has become the most famous and powerful method nowadays.
NWP is predicting the weather using a computer code of, governing equations, and
parameterizations of other physics processes. There are many studies about the use of
numerical weather prediction in research papers, paper articles, books, etc. Many countries
use NWP for various purposes. Aviation, navigation, flood forecasting, agriculture,
cyclones, heavy rainfall events are some examples. The main physics equations used in NWP
is shown in Table 2.1
Table 2.1 Physical equation of NWP
Type of equation Formula
Momentum F = ma
Mass 1 𝑑𝑀
=0
𝑀 𝑑𝑡
Moisture 𝑑𝑞
=𝐸−𝐶
𝑑𝑡
Energy 𝑑𝑇 𝑑𝛼
Q=Cv 𝑑𝑡 +p 𝑑𝑡
Ideal gas Pα= RT

2.2 General circulation models (GCMs)

A general circulation model (GCM) is a complex mathematical model that attempts to
provide information about the global climate. General circulation models (GCMs) are
essential tools for climate studies. The history of the general circulation modelling is broadly
divided into three phases characterized, respectively, by the rapid development of theories
on large-scale atmospheric motion, development of early GCMs, and expansion of internal
processes and use of higher resolutions. (Mechoso et al., 2015). The basic approximations,
boundary conditions, discretization of the governing equations, and the problem of
parameterization of physics processes are addressed.

4
2.3 Climate Downscaling
Downscaling is a method for obtaining high-resolution climate or climate change
information from relatively coarse-resolution global climate models (GCMs). Downscaling
techniques can be divided into two broad categories: dynamical and statistical. Dynamical
downscaling refers to the use of high-resolution regional simulations to dynamically
extrapolate the effects of large-scale climate processes to regional or local scales of interest.
Statistical downscaling encompasses the use of various statistics-based techniques to
determine relationships between large-scale climate patterns resolved by global climate
models and observed local climate responses. These relationships are applied to GCM results
to transform climate model outputs into statistically refined products, often considered to be
more appropriate for use as input to regional or local climate impacts studies. (Lanzante et
al., 2018) The strength and weakness of statistical vs dynamical downscaling are shown in
Table 2.2.

Table 2.2 Strength and weakness of statistical vs dynamical downscaling

Statistical downscaling Dynamical downscaling

Strength point scale climate information from Respond in physically consistent
GCM scale output ways to different external forcing
Cheap, computationally Resolve atmospheric processes such
undemanding, and readily as orographic precipitation
transferable
Ensembles of climate scenarios 10–50km resolution climate
permit risk/ uncertainty analyses information from GCM–scale output

Flexibility Consistency with GCM

Weakness Requires high-quality data for model Requires significant computing
calibration resources

Choice of predictor variables affects Initial boundary conditions affect

results results
Dependent on the realism of GCM Dependent on the realism of GCM
boundary forcing boundary forcing
Choice of domain size and location Choice of domain size and location
affects results affects results

2.3.1 Statistical downscaling

Typically, GCMs have a resolution of 150-300 km by 150-300 km. Many impact models
require information at scales of 50 km or less, so some method is needed to estimate the
smaller-scale information. Statistical downscaling first derives statistical relationships
between observed small-scale (often station level) variables and larger (GCM) scale
variables, using either analog methods (circulation typing), regression analysis, or neural
network methods. Future values of the large-scale variables obtained from GCM projections
of future climate are then used to drive the statistical relationships and so estimate the
smaller-scale details of future climate. (Wilby et al., 1998).
5
2.3.2 Dynamical downscaling

Dynamical downscaling involves the nesting of a higher resolution Regional Climate Model
(RCM) within a coarser resolution GCM. The RCM uses the GCM to define time-varying
atmospheric boundary conditions around a finite domain, within which the physics dynamics
of the atmosphere are modeled using horizontal grid spacings of 20–50 km. (Dawson et al.,
2004)

2.4 Dynamical downscaling model

There are multiple RCMS such as WRF, REGCM, MPI-REMO, UQAM-CRCM5 cite,
Further details of mostly used two models are given below.

2.4.1 REGCM model

REGCM is an open-source Regional Climatic Model developed by the Abdus Salam

International Center for Theoretical Physics, is used to downscale GCM (Mafas, 2016). The
dynamical component of the model originated from that of the MM4 (Mesoscale Model
Version 4) which is a compressible finite-difference model with hydrostatic balance and
vertical sigma coordinates. This included the Biosphere-Atmosphere Transfer Scheme
(BATS) for surface process representation, The Radiative Transfer scheme of the NCAR
Community Climate Model, A medium resolution local Planetary boundary layer scheme,
the Kuo-type cumulus convection scheme, and the explicit Moisture scheme.

2.4.2 WRF model

The Weather Research and Forecast model (WRF) is a numerical weather prediction and
atmospheric system designed for both research and operation application. (Mafas, 2016). It
is a supported “community model”, i.e., a free and shared resource with distributed
development and centralized support. WRF has two dynamical cores, The Advanced
Research WRF (ARW) and Nonhydrostatic Mesoscale Model (NMM). Its development is
led by NCAR, NOAA/ESRL, and NOAA/NCEP/EMC and collaborations with universities
and other government agencies in the US and overseas (Wang, 2020). There are so many
applications in the WRF model both research and weather prediction. Develop and test
physics parameterizations, Regional climate studies, Data assimilation research, and Real-
time (operational) forecasting are some examples of applications. WRF ARW has five steps
preprocessing system shown in Figure 2.1

6
 Figure 2.1 WRF ARW modeling system flow chart (Wang, 2020)

2.5 Physics scheme

Physics parameterizations approximate bulk effects of physical processes that are too small,
too brief, too complex, too poorly understood, or too computationally costly to be explicitly
represented. (Knievel, 2007). In the WRF Model, parameterizations include Microphysics
of clouds and precipitation, Radiation transferred through the atmosphere, Planetary
boundary layer and surface layer, Turbulence and diffusion, and Cumulus convection. Table
2.3 shows a summary of physics schemes that were used in studies conducted in Sri Lanka.

7
 Table 2.3 Summary of physics schemes which were used in research papers

Research Microphysics Cumulus Shortwave and Planetary Land

Paper Schemes Parameters Longwave Boundary surface
radiation layers model
schemes
Darshika Kessler, WSM5, Kain–Fritsch RRTM (LW), YSU Noah
D.W.T.T. WSM6 Betts-Miller– Dudhia (SW)
et al. Janjic
(2014)
Nandalal Thompson, Kain-Fritsch, Goddard s, Mellor Noah,
K.D.W. et Morrison, Betts-Miller- CAM, GFDL, Yamada, Rapid
al. (2012) WSM3, WSM6, Janjic, Grell- Dudhia, and YSU Update
Ferrier, Lin, et al Devenyi and RRTM Cycle
and Kessler New Grell (RUC)
and 5-layer
thermal
diffusion
model
Rodrigo C. WDM5, WDM6 Kain–Fritsch Dudhia and YSU Noah
et al. RRTM
(2018)
Weerakon Goddard Kain-Fritsch CAM(SW) Mellor- RCU
S.B. et al. and Yamada
(2016) RRTM(LW) Nakanishi
and Niino
Level

2.5.1 Microphysics scheme

The WRF 4.0 model has several optional microphysics options; Kessler scheme, Lin et al
scheme, WSM3 scheme (WRF Single Moment 3 class scheme), WSM5 scheme (WRF
Single Moment 5 class scheme), WSM6 scheme (WRF Single Moment 6 class scheme),
Ferrier scheme, Thompson scheme, and Morrison scheme. Kessler scheme (Kessler, 1969),
which was taken from the COMMAS (Collaborative Model for Multi-scale Atmospheric
Simulation) model (Wicker and Wilhelmson, 1995), is a simple warm cloud scheme that
includes water vapour, cloud water, and rain. Lin et al scheme has six classes of
hydrometeors including water vapour, cloud water, rain, cloud ice, snow, and graupel.
WSM3 scheme predicts three categories of hydrometeors: vapour, cloud water/ice, and
rain/snow, which is at times called a simple-ice scheme. WSM5 scheme is similar to the
WSM3 simple ice scheme. However, vapor, rain, snow, cloud ice, and cloud water are held
in five different arrays. WSM6 scheme is a six-class scheme that extends the WSM5 scheme
to include graupel and its associated processes. Ferrier scheme predicts changes in water
vapour and condensate in the forms of cloud water, rain, cloud ice, and precipitation ice
(snow/graupel/sleet). Thompson scheme is a new bulk microphysical parameterization
(BMP), which has been developed for use with the WRF model or other mesoscale models.
Compared to earlier single-moment BMPs the new scheme incorporates a large number of
improvements to both physical processes and computer coding. Morrison scheme is based
on a two-moment bulk microphysics scheme (Morrison et al., 2005). Six forms of water are
included: vapor, cloud droplets, cloud ice, rain, snow, and graupel/hail. The scheme
variables include several concentrations and mixing ratios of cloud ice, rain, snow, and

8
graupel/hail, and mixing ratios of cloud droplets and water vapor (total of 10 variables). The
prediction of two-moments (i.e., both the number of concentration and mixing ratio) allows
for a more robust treatment of the particle size distributions (Skamarock et al., 2008). Some
of the microphysics schemes used by models are illustrated in Figure 2.2.

Figure 2.2 Microphysics Parameterization Schemes. Mixing ratios Qv (water vapor), Qc

(cloud water), Qi (cloud ice), Qr (rain), Qs (snow), and Qg (graupel) are modeled. (Image
courtesy of Dudhia, 2004)

2.5.2 Cumulus scheme

The WRF model is armed with several options for the cumulus scheme, namely, Kain-
Fritsch, Betts-Miller-Janjic, Grell-Devenyi, and New Grell. The cumulus parameterization
scheme in the WRF model is one of the physics options in the model, which takes the effect
of cloud convection into account in predicting the weather. The direct concern of the
cumulus scheme is to predict convective precipitation. (Kuo et al., 1996)
Cumulus convection plays an important role in determining the behavior of weather and
global climate systems. Yet, for weather and climate prediction models the physical
processes associated with cumulus convection occur on scales that cannot be resolved by the
models. Representation of the effects of the cumulus convection in numerical models is
known as cumulus parameterization which is of fundamental importance in atmospheric
sciences (Nandalal, 2012)

9
2.5.3 Land surface scheme

Land surface scheme/models use atmospheric information from the surface layer scheme,
radiative forces from the radiation scheme, and precipitation from the microphysics and
convective schemes, together with internal information on the land’s state variables and land
surface properties, to provide heat and moisture fluxes over land points and sea-ice points.
Land surface models have various degrees of sophistication in dealing with thermal and
moisture fluxes in multiple layers of the soil and also may handle vegetation, root, and
canopy effects and surface snow cover prediction (Skamarock et al., 2008). The minimum
resolution requirements of the advanced land surface model are at least two soil layers, and
one canopy layer, this is because the top thin soil layer responds faster with the diurnal cycle
while the deeper layer controls the seasonal changes (Deardorff, 1977).

The Noah land surface model in the WRF model is a 4-layer soil temperature and moisture
model with canopy moisture and snow cover prediction. The layer thicknesses are 10, 30,
60, and 100 cm from the top to bottom. The total soil depth is 2 m, with the root zone in the
upper 1 m of soil. It includes the root zone, evapotranspiration, soil drainage, and runoff,
taking into account vegetation categories, monthly vegetation fraction, and soil texture
(Skamarock et al., 2008). It has one canopy layer and the following variables: soil moisture
and temperature in the soil layers, water stored on the canopy, and snow stored on the ground
(Xi, 2007). RUC land surface model has a multi-layer soil model (6 layers of soil) with
higher resolution in the top part of soil domain (0, 5, 20, 40, 160, 300 cm is the default)
(Smirnova et al., 1997). The RUC and Noah schemes are intermediate in detail and the major
difference between them is that RUC has a multi-layer snow scheme and Noah has a single
snow layer lumped with the topmost soil layer (Miller, 2007)

2.5.4 Longwave and shortwave radiation scheme

Similar to other weather models the WRF model takes the effects of radiation into account
in producing forecasts. The WRF model employs two types of separate radiation schemes,
one for shortwave radiation and the other for longwave radiation. It includes several
short/longwave radiation schemes that could be varied to see the impacts on precipitation
forecasts. For shortwave radiation schemes the WRF model contains Dudhia scheme (Model
default), Goddard scheme, CAM (Community Atmospheric Model) shortwave scheme, and
GFDL (Geophysical Fluid Dynamic Laboratory) shortwave scheme. Meanwhile, RRTM
(Rapid Radiative Transfer Model) scheme (WRF Model default), CAM scheme, and GFDL
schemes are available as longwave radiation options in the WRF model. (Nandalal et al.,
2012)

The radiation schemes provide atmospheric heating due to radiative flux divergence and
surface downward longwave and shortwave radiation for the ground heat budget. Longwave
radiation includes infrared or thermal radiation absorbed and emitted by gases and surfaces.
Upward longwave radiative flux from the ground is determined by the surface emissivity
that in turn depends upon the land-use type, as well as the ground (skin) temperature.
Shortwave radiation includes visible and surrounding wavelengths that make up the solar
spectrum. The processes include absorption, reflection, and scattering in the atmosphere and
at surfaces (Skamarock et al., 2008).

10
2.5.6 Planetary boundary layers scheme

The planetary boundary layer (PBL) which is, also called the atmospheric boundary layer,
is the region of the lower troposphere where Earth’s surface strongly influences temperature,
moisture, and wind through the turbulent transfer of air mass. As a result of surface friction,
winds in the PBL are usually weaker than above and tend to blow toward areas of low
pressure. (Hayden, 2016)

Planetary boundary layer (PBL) schemes are used to parameterize the vertical turbulent
fluxes of heat, momentum, and constituents such as moisture within the PBL as well as in
the free atmosphere. Each of these forces generated turbulence/eddies, which can be as deep
as the boundary layer itself. The accuracy of the PBL scheme is critical for forecasts of local
thermally and mechanically driven flows and air quality, and it also affects forecasts of
larger-scale meteorological phenomena (Hacker and Snyder 2005). Errors and uncertainties
associated with PBL schemes remain one of the primary sources of inaccuracies in model
simulations (Hu et al., 2010)

The WRF model includes NCEP GFS (Global Forecast System) scheme, ACM2
(Asymmetric Convective Model, version 2) scheme, MRF (Medium Range Forecast)
scheme, MellorYamada-Janjic (Eta) scheme, and Yonsei University (YSU) scheme under
its planetary boundary layer options. The WRF model includes NCEP GFS (Global Forecast
System) scheme, ACM2 (Asymmetric Convective Model, version 2) scheme, MRF
(Medium Range Forecast) scheme, Mellor-Yamada-Janjic (Eta) scheme, and Yonsei
University (YSU) scheme under its planetary boundary layer options. (Nandalal et al., 2012)

2.6 Status of WRF models in Sri Lanka

Rodrigo et al., 2018, assessed Sensitivity Study of WRF Numerical Modeling for
Forecasting Heavy Rainfall in Sri Lanka. This study focused on simulating two heavy
rainfall events observed over Sri Lanka using the WRF-ARW model. Four different
Microphysics schemes (WDM5, WDM6, WSM5, WSM6) and nine different Cumulus
parameterizations (Kain-Fritsch, Old Simplified Arakawa–Schubert, Betts–Miller–Janjic,
Grell–Freitas, Multi-scale KF, Grell-3, Tiedtke, New Tiedtke, New Simplified Arakawa–
Schubert) were used for this study.
Figure 2.3 shows, 24-h rainfall on 16 May 2016 for observation (a) model predictions
representing the model simulations using the selected (optimum configurations by graphical
representation) Microphysical and Cumulus parameterization schemes, (b) WSM6
Microphysics with BMJ cumulus only in the coarse domain, (c) WSM6 Microphysics with
BMJ cumulus in both the coarse and fine domains, (d) WDM5 Microphysics with Old SAS
cumulus in both the coarse and fine domains, (e) WSM6 Microphysics with Old SAS
cumulus in both the coarse and fine domains, and (f) WSM5 Microphysics with Old SAS
Cumulus in both the coarse and fine domains.

11
 Figure 2.3 Observed 24-h rainfall on 16 May 2016 and Model output

The overall results from the sensitivity tests of model horizontal resolution suggest that the
WRF-ARW runs using 3-km resolution were marginally better than those using 5-km
resolution. Simulations of heavy rainfall with the different physical parameterizations
showed significant sensitivity for both microphysics and cumulus schemes, but higher
sensitivity for cumulus schemes than microphysics.
The results of this study suggest that more studies of heavy rainfall cases are needed and that
testing more physics schemes, including radiation, boundary layer, and land surface physics,
to find better combinations of physics for heavy rainfall forecasting for all seasons would be
valuable. However, the quality of the WRF-ARW model predictions is highly dependent on
the lateral boundary and initial conditions provided for initialization. Thus, additional
sensitivity testing may be needed to determine the sensitivity of global model data such as
ECMWF.
Darshika, et al., (2014) simulated Heavy Rainfall From 19th to 28th December 2014 Using
WRF ARW (version3.0.1), for Different Atmospheric Physics in Sri Lanka. Two cumulus
parameterization schemes (Kain–Fritsch, and Betts–Miller–Janjic) and 3 microphysics
schemes (Kessler, WSM5, and WSM6) were used for this study. According to figure 2.4, all

12
the experiments except EXP2 indicate the heavy rainfall occurred in the eastern part of Sri
Lanka on 20th December 2014. All the experiments except EXP1 overestimated the rainfall
that occurred in the southwestern part of the island. The quantum of rainfall predicted in
EXP5 provides information about the rainfall distribution more accurately than other
experiments. However, all the experiments underestimated the rainfall distribution with
fairly widespread heavy rainfall that occurred over most parts of the island except in the
northern part on 25th December.

Figure 2.4 Simulated 24-hour cumulative rainfall (mm) for 20th Of December 2014 from
the WRF experiments and Observed rainfall data

This is a very preliminary attempt to use the WRF model (version 3.0.1), and it is
recommended that further testing using more case studies, as well as an improved version of
the model plus the incorporation of different satellite data and Doppler Radar data, through
use of the WRF data assimilation system, be carried out.
Application of WRF weather model to forecast precipitation in the Nilwala river basin was
done by Nandalal et al., (2012) Microphysics, Cumulus, Land surface, Boundary layer, and
Shortwave/ Longwave radiation Schemes were used for the study.
All the tested Microphysics schemes (Lin et al, Kessler, Thompson, Morrison, WSM3,
WSM6, and Ferrier) show a similar spatial distribution of accuracy over the basin for each
rain event. When it comes to the cumulus schemes a clear pattern of prediction accuracy
over the basin was not observed. The prediction accuracy changed spatially from event to
event as well as with different cumulus schemes used. The model default Kain-Fritsch
cumulus scheme produced reasonably good CPA% for all the three events. In the case of
Land surface options executed in model runs Noah, RUC, and Thermal diffusion models
have produced good predictions over the upper Nilwala river basin in all three rain events.
Figure 2.5 shows the Difference between WRF predictions and observed rainfall with
different Land surface model Mellor Yamada and YSU planetary boundary layer schemes

13
have shown very little influence on the CPA% and the spatial distribution of the accuracy of
the predictions.

Figure 2.5 Difference between WRF predictions and observed rainfall with a different
Land surface model
In this study, they have used only correctly predicted area percentage (CPA %) parameters
to visualize the goodness of predictions concerning observations. But there are more
evaluations such as Bias, mean absolute error, root means square error, etc. could be used
to improvise the results for this study. Also, by running the REG CM model to the same
study area, the results could be improved.
There are several research studies in weather research forecasting which were carried out
for Sri Lanka. But there is still no exact conclusion about physics schemes from the studies
in Sri Lanka and need further investigation and hope to fill that research gap by this study
to reasonable content.

14
2.7 Input data
Input data is very important for weather forecasting because all the outputs depend on the
input data. Mainly there are two types of input data. Metrological data and Geographical
data. Temperature, U, and V components of wind, surface pressure, relative humidity,
specific humidity are some examples for metrological data. Topographical height, soil data,
land use category, vegetation fraction are some Examples of geographical data. (Werner,
2017) the number of datasets that can be used as input to WPS can be downloaded directly
from the NCAR Research Data Archive (NCAR RDA) website (Rda.ucar.edu. 2020). Ocean
and Atmospheric Reanalysis, NCEP GDAS Observations and Analysis, NCEP GFS Model
Analysis and Forecast are popular data sets in NCAR. Table 3.1 shows the available GRIB
data sets from NCAR.

Table 2.4 Available GRIB data from NCAR

.

(Webmaster, M., 2020. WRF - Free Data. [online] Www2.mmm.ucar.edu.)

15
 CHAPTER THREE

MATERIALS AND METHODS

In this study, we are hoping to identify the best physics schemes for the Mi Oya basin
focusing on two heavy rainfall events (2016/05/13-2016/05/16 and 2019/12/18-2019/12/20)
which occurred in the Mi Oya basin. Figure 3.1 shows the methodology flow chart.

Figure 3.1 Methodology flow chart

3.1 Study area

The Mi Oya is a 108 km long river in North Western of Sri Lanka (shown in Figure 3.2). It
is the fifteenth- longest river in Sri Lanka. It begins in Saliyagama and flows northwest,
emptying into the Indian Ocean through Puttalam. Its catchment area receives approximately
2176 × 106 m3 of rain per year, and approximately 3% percent of the water reaches the sea.
It has a catchment area of 1516 km2

16
 Figure 3.2 Mi Oya river basin

3.2 Observation data Collection

Observed precipitation of Iginimitiya, Tabbowa irrigation, Puttalam, and Atharagalla
stations were collected from the meteorological department. This study hopes to use 24-
hour data from the above places, for 2019/12/18-2019/12/20 and 2016/05/13-2016/05/16
rainfall events.

Figure 3.3 Rainfall distribution over 24 hours on 06-11-2020

17
3.3 Methodology briefly

NCEP GDAS/FNL Reanalysis dataset is chosen as input data (shown in Table 2.4) for the
WRF ARW model.
The selection of optimum physics combination is important for accurate results. To identify
suitable schemes, A thorough literature review was carried out in Table 2.3 on previous
studies conducted in Sri Lanka. Specifically, six schemes were chosen. (shown in Table 3.1)
Table 3.1 Chosen physics schemes for the study

Microphysics schemes Kessler, WRF single moment 3 class

scheme (WSM3) and WRF single moment
class scheme (WSM6)

Cumulus scheme Kain–Fritsch and Betts–Miller–Janjic

schemes

Shortwave radiation scheme Dudhia scheme

Longwave radiation scheme RRTM scheme

Planetary Boundary layers YSU scheme

Land surface model Noah, and Rapid Update Cycle (RUC)

schemes

For this research, six combinations of physics schemes were chosen, shown in Table 3.2.
Then for analyzing WRF ARW output data ARC GIS software will be used. To compare
and validate the output data of each physics combination, four observational points of rainfall
data were chosen for the periods of rainfall events. After the comparison and validation, the
best physics combination can be identified for forecasting heavy rainfall in the Mi Oya river
basin.

18
 Table 3.2 Selected physics combinations

Combination 1 Kessler-Kain Fetish

Combination 2 Kessler-Betts miller Janic

Combination 3 WSM3-Kain Fetish

Combination 4 WSM6-Betts Miller Janic

Combination 5 WSM3-Betts Miller Janic

Combination 6 WSM6-Kain Fetish

3.4 Model configuration

In our research, we use Advanced Research WRF (ARW) version 4.0, hosted at the National
Center of Atmospheric Research (NCAR). The planetary boundary layer schemes, surface
layers, cumulus schemes, radiation schemes, and microphysics schemes are selected physics
schemes (illustrated in Table 3.1) for the study. A finer spatial resolution can improve the
quality of predictions, it tends to take a longer model run time to produce results and
demands superior computer resources for the simulations, which ultimately raises the
computation cost. Therefore, to obtain a healthy balance between accuracy and computation
cost, Nesting is used. Each integration is run with One-way nesting. The selected outer
domain is 27 km, the intermediate domain is 9 km and the inner domain is 3 km. The target
period of the accumulated forecasting rainfall is 24hrs. This is also the same as the 24hrs
rainfall criterion used in the Department of Metrology, Sri Lanka.

Figure 3.4 Selected WRF domain configuration in Weerakoon et al., (2016)

19
3.5 WRF Preprocessing system

The WRF Preprocessing System (WPS) is a set of three programs whose collective role is
to prepare input to the real program for real-data simulations. Each of the programs performs
one stage of the preparation, geogrid defines model domains and interpolates static
geographical data to the grids; ungrib extracts meteorological fields from GRIB-formatted
files; and metgrid horizontally interpolates the meteorological fields extracted by ungrib to
the model grids defined by geogrid. The work of vertically interpolating meteorological
fields to WRF eta levels is performed within the real program. Figure 3.5 shows the weather
research forecasting pre-processing system.

Figure 3.5 WPS Flow diagram (Diagne et al.,2013)

3.5.1 Geogrid

The purpose of geogrid is to define the simulation domains and interpolate various terrestrial
data sets to the model grids. The simulation domains are defined using the information
specified by the user in the “geogrid” name list record of the WPS name list file, name list
WPS. In addition to computing the latitude, longitude, and map scale factors at every grid
point, geogrid will interpolate soil categories, land use category, terrain height, annual mean
deep soil temperature, monthly vegetation fraction, monthly albedo, maximum snow albedo,
and slope category to the model grids by default.
3.5.2 Ungrib

The ungrib program reads GRIB files, "degribs" the data, and writes the data in a simple
format called the intermediate format. The GRIB files contain time-varying meteorological
fields and are typically from another regional or global model, such as NCEP's NAM or GFS
models.

20
3.5.3 Metgrid

The metgrid program horizontally interpolates the intermediate-format meteorological data

that are extracted by the ungrib program onto the simulation domains defined by the geogrid
program. The interpolated metgrid output can then be ingested by the WRF real program.
WRF Standard Initialization (2020)

3.6 Validation methods

Verification will be carried out for the 3 km domain and 24 h accumulated rainfall at 24 h
forecast ranges. After the model was run using the different parameterization combinations,
the total model performance can be calculated by first calculating the bias, MAE, RMS,
Correlation Coefficient (corr), and slope of each model output data in the particular grid (The
grid where the particular rainfall station located) concerning each observed rainfall station
data. To determine the overall model performance, these metrics would have to be combined.
Since the bias, MAE, and RMS have units, they cannot be added right away to the correlation
and slope. They can be converted to dimensionless values by dividing the bias, MAE, and
RMS by their respective observation means. They then added to |1-corr| and to |1-slope| and
the average can be calculated. This is the total model performance or TMP. It ranges from 0
to 1 with 0 indicating that the model output value is close to the observations and 1 indicating
that the model output value is far from observations.
Table 3.3 summarized the equation of bias, mean absolute model error (MAME), root mean
square error (RMSE), slope, correlations coefficient, and Total model performance (TMP)
will be used for evaluating model performance in comparison to observations. Ym is the
model output, Yo is the observation, Yo, mean is the mean of the observations and N is the total
number of data points.

21
 Table 3.3 Summary of evaluation equations

Evaluation method Equation

𝑛
Bias 1
∑(𝑌𝑚 − 𝑌𝑜)
𝑁
𝑖=1
𝑛
Mean absolute error 1
(MAE) ∑ |𝑌𝑚 − 𝑌𝑜|
𝑁
𝑖=1
Root mean square 𝑛
error (RMSE) 1
√ ∑(𝑌𝑚 − 𝑌𝑜)2
𝑁
𝑖=1

Slope Linear Regression

Correlations Coefficient Pearson product-moment

Total Model Performance 𝑀𝐴𝐸 + 𝑅𝑀𝑆 |1 − 𝑆𝑙𝑜𝑝𝑒|

(TMP) + (1 − 𝐶𝑜𝑟𝑟) +
Yo, (mean) 4

The major outcomes of the research are to obtain an optimum physics combination to
forecast heavy rainfall in the Mi Oya river basin and provide a preliminary study for future
numerical forecasting studies

22
 CHAPTER 4

4.1 Preliminary studies

The majority of the WRF model codes are written in the Fortran language. To run this
model a higher performance computer is required, so we connected to a Linux server
(provided by the project supervisor) using SSH terminal emulator called MobaXterm
(Figure 4.1). Then, basic Linux commands were practiced.

Figure 4.1 MobaXterm SSH terminal emulator

23
Work schedule

24
 Insert Project Title and project number Here
Problem Background & Find the numerical weather Prediction capabilities during heavy
Project Description: rainfall. Especially focused on Mi Oya river basin

Project Objective(s): • Set up the WRF model

• Find optimum physics scheme for Mi Oya basin
a bullet list of clear objectives

Financial Requirement: • Budget: Rs. 560

Industrial Partners
(if applicable):

Expected Project • Poster

Deliverables:
(a bullet list of clear
deliverables)

Ethical and Data Please tick if any of the following apply.

Protection Issues □ The project has data protection issues; please specify.
□ The project has ethical issues; please specify.
□ The project involves human subjects; please specify.

Students Names: GAMAGE K.H. (E/15/107)

WANASINGHE W.M.A.Y (E/15/376)

Supervisors Name: Dr. N.G.P.B. NELUWALA

Area of Study: Numerical weather forecasting

Panel number: C/15/58

SIGNATURES (Students):

SIGNATURE/S (Supervisor/s):

DATE:
 REFERENCES

1. Nandalal, K., Sachindra, D. and Ratnayake, U. (2012). Application of WRF weather

model to forecast precipitation in the Nilwala river basin. Engineer: Journal of the Institution
of Engineers, Sri Lanka, 45(1), pp.51-67.
2 Mafas M.M.M., Muhammadh K.M., Weerakoon S.B. and Mutua F. (2016) Comparative
study of WRF & REGCM weather predictions for the upper Mahaweli river basin.1.
3. Evans, J.P., Ekström, M. & Ji, F. (2012). Evaluating the performance of a WRF physics
ensemble over South-East Australia. Clim Dyn 39, pp. 1241–1258.
https://doi.org/10.1007/s00382-011-1244-5
4. van Vuuren, D., Edmonds, J., Kainuma, M., Riahi, K., Thomson, A., Hibbard, K., Hurtt,
G., Kram, T., Krey, V., Lamarque, J., Masui, T., Meinshausen, M., Nakicenovic, N., Smith,
S. and Rose, S. (2011). The representative concentration pathways: an overview. Climatic
Change, 109(1-2), pp.5-31.
5. Lanzante, J., Dixon, K., Nath, M., Whitlock, C. and Adams-Smith, D. (2018). Some
Pitfalls in Statistical Downscaling of Future Climate. Bulletin of the American
Meteorological Society, 99(4), pp.791-803.
6.Wilby, R., Dawson, C., and Barrow, E. (2002). sdsm a decision support tool for the
assessment of regional climate change impacts. Environmental Modelling & Software,
17(2), pp.145-157.
7. Pleim, J. E. (2007). A combined local and nonlocal closure model for the atmospheric
boundary layer. Part I: Model description and testing.J. Appl. Meteor. Climatol., 46, pp.
1383–1395

8. Wilby, R., Wigley, T., Conway, D., Jones, P., Hewitson, B., Main, J. and Wilks, D. (1998).
Statistical downscaling of general circulation model output: A comparison of methods.
Water Resources Research, 34(11), pp.2995-3008.
9. Roger A. Pielke (2013). Chapter 13 - Examples of Mesoscale Models, Editor(s): Roger
A. Pielke, International Geophysics, Academic Press, Volume 98, pp. 427-500, ISSN 0074-
6142, ISBN 9780123852373, <https://doi.org/10.1016/B978-0-12-385237-3.00013-X.>
10. NCAR's RDA. (2020). Available at: <https://rda.ucar.edu/> [Accessed 28 November
2020] Rda.ucar.edu. 2020. NCAR's RDA. [online] Available at: <https://rda.ucar.edu/>
[Accessed 28 November 2020]
11. Powers, J. G., and Coauthors. (2017). The Weather Research and Forecasting Model:
Overview, System Efforts, and Future Directions. Bull. Amer. Meteor. Soc., 98, pp.1717–
1737, <https://doi.org/10.1175/BAMS-D-15-00308.1>

26
12. Oreskes, N. (2004). The Scientific Consensus on Climate Change. Science, 306(5702),
pp.1686-1686.
13. Channa Rodrigo, Sangil Kim, and Hyo Jung. (2018). Sensitivity Study of WRF
Numerical Modeling for Forecasting Heavy Rainfall in Sri Lanka, Atmosphere (ISSN 2073-
4433; CODEN: ATMOCZ)
14. Darshika, D.W.T.T. and Premalal, K.H.M.S. (2014). Simulate Heavy Rainfall During
19th to 28th December 2014 Using WRF for Different Atmospheric Physics,
<http://www.meteo.gov.lk/images/sljom/thanuja.pdf>
15. Www2.mmm.ucar.edu. (2020). Chapter 3: WRF Standard Initialization. [online]
Available at: <https://www2.mmm.ucar.edu/wrf/users/docs/user_guide_V3/users_guide_
chap3.htm> [Accessed 28 November 2020]
16. Scied.ucar.edu. (2020). What Is Weather? | UCAR Center For Science Education.
[online] Available at: <https://scied.ucar.edu/learning-zone/how-weather-works/weather>
[Accessed 27 November 2020]

17. Anil Kumar, R., Dudhia, J., & Bhowmik, S. K. (2010). Evaluation of Physics options of
the Weather Research and Forecasting Model to simulate high impact heavy rainfall events
over the Indian Monsoon region. Geofizika, 27(2), pp. 101-125.

18.Www2.mmm.ucar.edu. (2020). [online] Available at: <https://www2.mmm.ucar.edu

/wrf/users/tutorial/202001/wang_intro.pdf> [Accessed 23 November 2020]

20. National Weather Service (2020). [online] Available from:<https://www.weather.gov/

[25 November 2020]>

21. Febri, D., Hidayat, R. and Hanggoro, W. (2015). The sensitivity of WRF-EMS model to
predict rainfall events in wet and dry seasons over West Sumatra. Procedia Environmental
Sciences, 33, pp.140-154.

22. Webmaster, M. (2020). WRF - Free Data. [online] Www2.mmm.ucar.edu. Available at:
<https://www2.mmm.ucar.edu/wrf/users/download/free_data.html>
[Accessed 28 November 2020]

23.Large W.G. (1998) Modeling and Parameterizing the Ocean Planetary Boundary Layer.
In: Chassignet E.P., Verron J. (eds) Ocean Modeling and Parameterization. NATO Science
Series (Series C: Mathematical and Physical Sciences), vol 516. Springer, Dordrecht.
https://doi.org/10.1007/978-94-011-5096-5_3

24.Hu, X., Nielsen-Gammon, J., and Zhang, F., (2010). Evaluation of Three Planetary
Boundary Layer Schemes in the WRF Model. Journal of Applied Meteorology and
Climatology, 49(9), pp.1831-1844.

25.Hacker, J. P., and C.Snyder. 2005. Ensemble Kalman filter assimilation of fixed screen-
height observations in a parameterized PBL.Mon. Wea. Rev., 133, pp. 3260–3275

27