GLM v2 User Manual PDF

2014
Oct
- General Lake Model
Model Overview and User Information

v 2.0 β - Oct 2014
M.R. Hipsey, L.C. Bruce & D.P. Hamilton

Summary
The General Lake Model (GLM) is an open-access model developed for simulating lake dynamics. It simulates
vertical stratification and mixing and accounts for the effect of inflows/outflows, surface heating and cooling,
and it can be extended to include the effect of ice cover. GLM has been designed to be an open-source
community model developed in collaboration with members of Global Lake Ecological Observatory Network
(GLEON) to integrate with lake sensor data.
It is suited to environmental modelling studies where simulation of lakes or reservoirs is required. The one-
dimensional (1D) basis of the model means it is suited to seasonal and decadal scale investigations of water
quality but it can also be used in comparisons of simulation output against high-frequency sensor data. Sites
that may be simulated with the model include deep and shallow lakes, drinking water, hydropower or
irrigation reservoirs, mining pit lakes, wastewater ponds and urban wetlands. The model couples with the
Aquatic EcoDynamics library (AED) for integrated simulations of lake and reservoir water quality and
ecosystem health.
This manual summarises the scientific basis and numerical implementation of the model algorithms including
the sub-models related to surface heat exchange and ice-cover dynamics, vertical mixing and the
inflow/outflow dynamics. A summary of typical parameter values for lakes and reservoirs collated from a range
of sources is included. The final section provides an overview of setting up and running the model. Further
information for analysis of model outputs and undertaking sensitivity and uncertainty assessments with the
model is also provided.
Acknowledgements
GLM has been developed by the Aquatic EcoDynamics Research Group team at UWA (http://aed.see.uwa.edu.au) in collaboration with
the Lake Ecosystem Restoration research group at the University of Waikato (http://www.lernz.com). Funding for the initial development
of the model was from the U.S. NSF Cyber-enabled Discovery and Innovation (CDI) grant awarded to Prof. Paul Hanson (lead investigator)
and colleagues from 2009-2014. In addition to the named authors who have driven the scientific direction of the model, we acknowledge
the software engineering of the model undertaken by Casper Boon at UWA. We also acknowledge numerous members of the GLEON
ecosystem modelling working group and the associated multi-lake comparison project (MLCP), and in particular are grateful to Jordan
Read and Luke Winslow form the USGS for extensive help. Through application of the model within the GLEON community we have been
able to undertake extensive review of model performance on earlier versions and gain essential feedback on model features and
operation. We also acknowledge development of the Bird Clear Sky Model for use within GLM by Aditya Singh from UWA.
Whilst GLM is a new code, it is based on the large body of historical research and publications produced by the Centre for Water
Research at the University of Western Australia, which we acknowledge for the inspiration and guidance on the approach that has been
adopted herein . We also acknowledge the MCMC code by Marko Laine that has been integrated with this GLM version model sensitivity
and uncertainty assessment, (accessed from http://helios.fmi.fi/~lainema/mcmc/). Provision of the environmental symbols used for the
GLM scientific diagrams are courtesy of the Integration and Application Network, University of Maryland Center for Environmental
Science.
Citing this manual:

Hipsey, M.R., Bruce, L.C., Hamilton, D.P., 2014. GLM - General Lake Model: Model overview and user information. AED Report #26, The
University of Western Australia, Perth, Australia. 42pp.
Copyright © 2014. The University of Western Australia.

ISBN: 978-1-74052-303-5
2
aed.see.uwa.edu.au/research/models/GLM
Oct 2014
Contents
OVERVIEW ...................................................................................................................................................................... 5
BACKGROUND 5
MODEL APPROACH 6
MODEL SUITABILITY & DATA REQUIREMENTS 6
MODEL SCIENCE OVERVIEW ..................................................................................................................................... 7

LAYER STRUCTURE 7
WATER BALANCE 7
SURFACE ENERGY BUDGET 8
Solar heating and light penetration 8
Longwave radiation 10
Sensible and latent heat transfer 11
Correction for non-neutral atmospheric stability .................................................................................................................. 11
Still-air limit ............................................................................................................................................................................... 11
Wind-sheltering ....................................................................................................................................................................... 12
SNOW AND ICE DYNAMICS 12
STRATIFICATION AND VERTICAL MIXING 14
Surface Mixed Layer: 14
Deep Mixing: 16
INFLOWS AND OUTFLOWS 17
River inflows 17
Submerged inflows 18
Withdrawals 18
Seepage 19
BOTTOM STRESS 19
SETUP & OPERATION ............................................................................................................................................... 23

OVERVIEW 23
INPUT FILES 23
Physical model configuration: glm.nml ........................................................................................................................... 23
Meteorological configuration and met.csv ..................................................................................................................... 24
Configuration of inflows and setup of inflows.csv .......................................................................................................... 26
Configuration of outflows and setup of outflows.csv .................................................................................................... 27
Configuring the model to run water quality ................................................................................................................... 27
RUNNING THE MODEL 28
OUTPUTS AND POST-‐PROCESSING 28
Live output plotting: plots.nml ........................................................................................................................................ 28
Simulation summary: lake.csv .......................................................................................................................................... 29
Plotting in EXCEL .............................................................................................................................................................. 30
Plotting in MATLAB .......................................................................................................................................................... 30
Plotting in R ....................................................................................................................................................................... 30
MODEL VALIDATION & PARAMETER OPTIMISATION 31
Running the LakeAnalyzer validation .............................................................................................................................. 31
Running the MCMC parameter estimation .................................................................................................................... 31
3
Oct 2014

EXAMPLES & SUPPORT ........................................................................................................................................... 34

DOWNLOADS & FURTHER SUPPORT 34
EXAMPLE APPLICATIONS 34
REFERENCES ............................................................................................................................................................... 35
APPENDICES ............................................................................................................................................................... 38

A: BIRD SOLAR RADIATION MODEL 38
B: NON-‐NEUTRAL BULK TRANSFER COEFFICIENTS 40
4
Oct 2014
Overview
Background
The General Lake Model (GLM) is a one-dimensional hydrodynamic model for simulating the water balance and vertical
stratification of lakes and other standing (lentic) water bodies. GLM computes vertical profiles of temperature, salinity and
density by accounting for the effect of inflows and outflows on the water balance, in addition to surface heating and
cooling, and vertical mixing (Figure 1). The model also includes the effects of ice cover formation and subsequent melting
on the heating and mixing processes within the lake (Yao et al., 2014).
Since the model is one-dimensional it assumes no horizontal variability within the domain and users must therefore ensure
their application of the model is suited to this assumption. For deep, stratified, systems, the model is ideally suited to
long-term investigations ranging from months to decades, and for coupling with biogeochemical models to explore the
role that stratification and vertical mixing play on lake ecosystem dynamics. However, the model can also be used, with
some caution and checks, for shallow lakes, ponds or wetlands where the water column is relatively well mixed. A recent
application of the model demonstrates its ability for including lakes in regional climate and earth system assessments
(Read et al., 2014).
The model was initially built as a project within the Global Lake Ecological Observatory Network (GLEON) to
provide a computationally efficient lake modelling platform to be used for integration with lake observatory systems and
for training lake scientists. The model couples with the Framework for Aquatic Biogeochemical Models (FABM),
and in particular is designed to operate with the Aquatic EcoDynamics modules (Hipsey, 2014) included within FABM
(termed FABM-AED). Since its original development, the model has also been used successfully for simulating reservoirs,
mining pit lakes and wetlands. The model is available freely and distributed as open-source under the GNU GPL license,
and it is encouraged that the tool be adapted for a wide variety of applications so that we can advance lake simulation
(Mooij et al., 2010; Trolle et al., 2012).
Figure 1: Schematic of a GLM simulation domain, input information (blue text) and
key simulated processes (black text).
5
Oct 2014
Model Approach
For background information of on the mixing dynamics of lakes readers are referred to summaries by Imboden and Wüest
(1995) and Imberger and Patterson (1990). GLM adopts a one-dimension solution processes of vertical mixing by
incorporating a series of vertical layers that are used to describe the variation in water column properties. The model
adopts a flexible Lagrangian structure originally introduced for the model DYRESM by Imberger et al. (1978) and Imberger
& Patterson (1981). Numerous model variations have since been introduced to further extend this conceptual approach
through applications to a variety of lake and reservoir environments (e.g., Hocking & Patterson, 1991; Hamilton &
Schladow 1997; McCord & Schladow, 1998; Gal et al., 2003; Yeates et al., 2004). The Lagrangian design assumes each layer
is a ‘control volume’ that can change thickness by contracting and expanding in response to inflows, outflows, mixing with
adjacent layers, and surface mass fluxes. Layers each have a unique density computed based on the local salinity and
temperature and when sufficient energy becomes available to over come density differences between the adjacent layers,
they will merge thus accounting for the process of mixing. For deeper systems, a stable vertical density gradient will form
in response to periods of high solar radiation creating warm, less-dense conditions near the surface with cooler conditions
deeper in the water separated by a thermocline region (metalimnion). Layer thicknesses are adjusted throughout the water
column by the model in order to sufficiently resolve the vertical density gradient with fine resolution occurring in the
thermocline and thicker cells where mixing is occurring (as depicted schematically in Figure 1). Unlike the fixed grid
(Eulerian) design of most lake and ocean models, where mixing algorithms are typically based on resolving vertical
velocities, it has been reported that numerical diffusion of the thermocline in this approach is limited, making it particularly
suited to long-term investigations, and requiring limited site-specific calibration (Patterson et al., 1984; Hamilton &
Schladow, 1997).
Although GLM is a new model code written in C, the core layer structure and mixing algorithms have been based on
equations summarised in Hamilton and Schladow (1997) and Imberger and Patterson (1981), thereby making it similar to
these previously reported models. Beyond this functionality, the model features numerous customisations and extensions
in order to make it a fast and easy to use package suitable for a wide range of contemporary applications.
Model Suitability & Data Requirements

The model may be suitable for investigations where resolving the horizontal variability is not a requirement of the study.
This includes natural lakes, large and small drinking water reservoirs and mining pit lakes. Despite the 1D assumption, the
model performs well in reservoirs with dendritic morphometry (e.g., drowned river valleys) or more regular shapes. Whilst
the model is able to resolve vertical stratification, it may be used to simulate shallow lakes, wetlands, wastewater ponds
that experience well-mixed conditions. In this case, the layer structure will automatically simplify, and mass and energy will
continue to be conserved. The model is suitable for operation in a wide range of climate conditions and is able to simulate
ice formation, and also accommodate a range of atmospheric conditions.
Beyond modelling the water and heat balance of a lake, the model is well-suited to simulate water quality investigations
through coupling with a water quality model library. The model is distributed pre-compiled with the AED WQ modelling
libraries, and these are able to simulate turbidity, oxygen, nutrients, phytoplankton, zooplankton, pathogens and chemical
variables.
In general, the model can be simplified to essential a body of water with minimal complexity. Users may configure any
number of inflows and outflows and more advanced options exist for simulating the water and heat balance. At a
minimum, the model requires the user to supply a hypsographic curve, A = A(h), to describe the storage, elevation, and
area relationships. Depending on the context of the simulation, either daily our hourly meteorological time-series data for
surface forcing is required, and daily time-series of volumetric inflow and outflow rates may also be required. A summary
of relevant parameters within the model and their default values are given in Tables at the end of the model science
overview section.
The model may be run without any 3rd party software as the input files consist of “namelist” text files for configuration and
csv files for meteorological and flow data. Further details of the model setup and file formats are outlined in the GLM
Setup section.
6
Oct 2014
Model Science Overview
Layer structure
The model is composed of a series of layers numbered from the lake bottom to the surface. The number of layers, 𝑁!"# (𝑡),
is adjusted throughout the simulation to maintain the assumption that each layer must have homogenous properties
across the layer. Initially, the layers are assumed to be of equal thickness, and the initial number of layers, 𝑁!"# (𝑡 = 0),
depends on the user-defined minimum (ℎ!"# ) and maximum (ℎ!"# ) layer thickness limits that are set, and the lake depth
(both defined in glm.nml, see model setup section). As the model simulation progresses, density changes due to surface
heating, vertical mixing, and inflows and outflows lead to dynamic changes in the layer structure as layers expand or
contract. The model includes routines to enforce the layer limits, maintaining the optimal thickness of layers required to
resolve the vertical density gradient.
The layer volumes are determined by interpolating layer area off the user-specified hypsographic curve for the lake basin,
such that 𝐴! = 𝑓(ℎ! ), where 𝑖 is the layer number. The user provides 𝑁!"# depth points with basin area to define the
hypsographic curve. Layers are generally at a relatively coarse resolution relative to the simulated layers, and the model
can either i) accept prescribed volume values at each, or ii) compute the volumes assuming a simple interpolation. In the
latter case, the first layer, 𝑉! , is computed assuming a conical shape, and above that each point as:
𝑉! = 𝑉!!! + 𝐴!!! + 0.5(𝐴! − 𝐴!!! ) (ℎ! − ℎ!!! ) (1)
where 1 < 𝑏 ≤ 𝑁!"# . Using the raw hyposgraphic data, a refined depth-area-volume relationship is calculated during the
simulation using finer depth increments (e.g., ~ 0.1 m), giving 𝑁!"#$% levels that are used for subsequent calculations. The
area and volume at the depth of each increment, ℎ! is interpolated from the supplied information as:
!! !! !! !!
𝑉! = 𝑉!!! and 𝐴! = 𝐴!!! (2)
!!!! !!!!
where 𝑉! and 𝐴! are the volume and area at each of the refined elevations of the refined depth vector, and 𝑉! in these
expressions refers to the nearest b level below ℎ! such that ℎ!!! < ℎ! . Note the interpolation coefficients are computed as:
!!!! ! !!!
!"#!" !! !"#!" !!
𝛼! = ! and 𝛽! = !!!! (3)
!"#!" !!!
!!
!"#!" !!
The density in each layer is computed based on the temperature, 𝑇, and salinity, 𝑆, at any given time according to the
UNESCO (1981) equation of state: 𝜌! = 𝜌 𝑇! , 𝑆! .
Water balance
The model solves the water balance of the lake domain by including several user-configurable fluxes. A daily summary of
the water balance is provided to the user via the summary information in lake.csv. The main water balance components
include:
• Surface mass fluxes
o Evaporation
o Rainfall
o Snowfall
• Inflows
o Surface inflows
o Deep inflows
o Runoff from the surrounding catchment
• Outflows
o Withdrawals
o Overflow
o Seepage
7
Oct 2014
The mass balance occurs through the layers, but evaporation and precipitation only occur in the surface layer and seepage
only occurs form the bottom layer. The model computes the dynamics of the inflows and outflows on a daily time-step,
however the surface mass fluxes can occur hourly or daily depending on the resolution of meteorological forcing data.
The change in surface layer thickness due to surface mass fluxes is summarised as:
𝑑ℎ!
= 𝐸 + 𝑆 + 𝑓! 𝑅 + 𝑄! 𝐴! (4)
𝑑𝑡
where hS is the height of the surface layer (m), t time step (s), E is the evaporation mass flux computed from the heat flux
𝜙! (W m-2) described below, R is rainfall and S is snowfall (m day-1), and 𝑓! is a user-definable scaling factor that may be
applied to increase or reduce the rainfall data (default = 1). 𝑄! is an optional term to account or runoff to the lake form
the exposed banks, which may be important in reservoirs with a large drawdown range, or wetlands where periodic drying
of the lake may occur. This calculated using a simple runoff model when the rainfall intensity exceeds the threshold, RL (m
day-1):
𝑄! = 𝑓!" 𝑓! 𝑅 − 𝑅! 𝐴!"# − 𝐴! (5)
where 𝑓!" is the runoff coefficient, defined as the fraction of rainfall that is converted to runoff to the lake’s edge, and ius
the maximum possible area of inundation of the lake (as defined by the area provided by the user at 𝑁!"# area value).
Note that Eq 4 does not include changes to hS as a result of mixing dynamics (i.e. the merging or splitting of layers ot
enfoce the layer thickness limits), or ice formation/melt, or river inflows; these are described in subsequent sections.
However, all layers within the domain also are subject to mass conservation and impacted by inflowing and outflowing
water.
Surface energy budget

A balance of shortwave and long wave radiation fluxes, sensible and evaporative heat fluxes determine the net cooling
and heating for GLM. The general heat budget equations can be described as:
𝑐! 𝑑𝑇!
= 𝜙!"! − 𝜙! + 𝜙! + 𝜙!"#$ − 𝜙!"#$% (6)
𝐴! 𝑧!"# 𝑑𝑡
where cp is the specific heat capacity of air (1005 J/kg/°C), 𝑇! is the surface temperature of the surface mixed layer and the
RHS heat flux terms are expanded upon individually below. Several options exist for customizing the individual surface
heat flux components and also summarised below.
Solar heating and light penetration

Solar radiation is the key driver of the lake thermodynamics, however, data may not always be available from a nearby
pyranometer. GLM v2 users may choose to either have GLM compute surface irradiance from a theoretical approximation
based on the Bird Clear Sky insolation model (Bird, 1984), modified for cloud cover, or alternatively hourly or daily solar
radiation intensity data may be directly specified. If the former, then 𝜙!" is calculated from (Bird, 1984; Luo et al., 2010):
𝜙!" + 𝜙!"
𝜙!" = 𝑓(𝐶) (7)
1 − 𝛼!" 𝛼!"#
where the model computes total irradiance, 𝜙!" (W m-2), from direct beam 𝜙!" , and the irradiance from atmospheric
scattering 𝜙!" (refer to Appendix A for a detailed outline of the BCSM equations and parameters). In GLM, the clear sky
value is reduced according to the amount of clouds, C, according to:
𝑓 𝐶 = 0.66182 𝐶 ! − 1.5236 𝐶 + 0.98475 (8)
which is based on a regression of cloud data from Perth Airport, compared against nearby sensor data (R2 = 0.952; see also
Luo et al., 2010).
8
Oct 2014
Figure 2: Schematic of a surface heat fluxes impacting the surface mixed layer (sml).
The albedo, 𝛼!" , is the reflected fraction of 𝜙!" , with several options via radmode in glm.nml:
Option 1 : Daily approximation, Hamilton & Schladow (1997)
!! !
0.08 + 0.02 𝑠𝑖𝑛 𝑑− ∶ 𝑛𝑜𝑟𝑡ℎ𝑒𝑟𝑛 ℎ𝑒𝑚𝑖𝑠𝑝ℎ𝑒𝑟𝑒
!"# !
(9a)
𝛼!" = 0.08 ∶ 𝑒𝑞𝑢𝑎𝑡𝑜𝑟
!! !
0.08 − 0.02 𝑠𝑖𝑛 𝑑− ∶ 𝑠𝑜𝑢𝑡ℎ𝑒𝑟𝑛 ℎ𝑒𝑚𝑖𝑠𝑝ℎ𝑒𝑟𝑒
!"# !
Option 2 : Briegleb et al., (1986)
1 2.6 (9b)
𝛼!" = + 15 𝑐𝑜𝑠 Φ!"# − 0.1 𝑐𝑜𝑠 Φ!"# − 0.5 𝑐𝑜𝑠 Φ!"# − 1
100 1.1𝑐𝑜𝑠 Φ!"# !.! + 0.065
Option 3 : Yajima & Yamamoto (2014)
!.!! !!.!" !.!"# (9c)

𝛼!" = 0.001 RH cos Φ!"# − 0.001 𝑈! cos Φ!"# − 0.001 ς cos Φ!"#
where d is the day of the year, and Φ!"# is the solar zenith
angle (radians) as outlined in Appendix A, RH is the relative 0.4
Briegleb et al. (1986)
humidity and ς is the atmospheric diffuse radiation. The 2 nd
0.35
(oceanic) and 3rd (lacustrine) options allow for diel and Yajima & Yamamoto (2014)
0.3
seasonal variation of albedo from approximately 0.01 to 0.4
depending on the sun-angle (Figure 3). 0.25
Albedo
0.2
Shortwave radiation penetration into the lake and through 0.15
the layers is modelled according to the Beer-Lambert Law:
0.1
𝜙!" 𝑧 = 1 − 𝛼!" 𝑓!" 𝑓!"# 𝜙!" 𝑒𝑥𝑝 −𝐾! 𝑧 (10)
0.05
where z is the depth of the layer from the surface, 𝑓!" is a 0

scaling factor that may be applied and adjusted as part of 90 80 70 60 50 40 30 20 10 0
the calibration process, and Kw is the light extinction SZA
coefficient (m ). Kw may be set by the user as constant or
-1
Figure 3: Variation o f albedo (𝜶𝑺𝑾 ) with s olar z enith
linked to the water quality model (e.g. FABM or AED2) in angle (SZA = 𝟐𝛑 𝚽𝒛𝒆𝒏 ⁄𝟏𝟖𝟎, deg rees ) for radmode 2 an d 3.
9
Oct 2014
which case the extinction coefficient will change as a function of depth and time according to the dissolved and particulate
constituents. In the model, Beer’s Law is only applied for the photosynthetically active fraction (PAR), 𝑓!"# , which is 45% of
the incident light. The amount of light heating the surface layer, 𝜙!"! , is therefore the above photosynthetically average
fraction that enters across zSML, plus the remaining 1 − 𝑓!"# fraction which accounts for the fact that near infra-red and
ultraviolet bandwidths of the incident shortwave tend to have significantly higher attenuation coefficients (Kirk, 1994).
In some applications the extent to which the benthos has a suitable light climate is a good indicator of benthic
productivity, and a proxy for the type of benthic habitat that might emerge. GLM predicts the benthic area of the lake that
exceeds a user defined light intensity, 𝜙!"#!"#$ .
𝐴!"# = 𝐴! − 𝐴 ℎ!"# (11)
where ℎ!"# = ℎ!"#$ − 𝑧!"# , and 𝑧!"# is calculated from Beer’s law:
𝜙!"#!"#$ −1
𝑧!"# = 𝑙𝑛 (12)
𝜙!"! 𝐾!
The daily average benthic area is reported in the lake.csv summary file as a percentage (𝐴!"# /𝐴! ).
Longwave radiation
Longwave radiation can either be specified as net flux, incoming flux or if there is no radiation data from which longwave
radiation can be computed, then it may be calculated by the model internally based on the cloud cover fraction and air
temperature. Net long wave radiation is described as:
𝜙!"!"# = 𝜙!"!" − 𝜙!"!"# (13)
where
!
𝜙!"!"# = 𝜀! 𝜎 𝑇! + 273.15 (14)
and σ is the Stefan-Boltzman constant and εw the emissivity of the water surface, assumed to be 0.985. If the net or
incoming longwave flux is not provided, the model will compute the incoming flux from:
𝜙!"!" = 1 − 𝛼!" 𝜀!∗ 𝜎 𝑇! + 273.15 !

(15)
where 𝛼!" is the long-wave albedo (0.03), and the emissivity of the atmosphere is computed considering emissivity of
cloud-free conditions (𝜀! ), based on air temperature and humidity, extended to account for reflection from clouds, such
that 𝜀!∗ = 𝑓 𝑇! , 𝐶 calculated from (Henderson-Sellers, 1986):
1 + 0.275𝐶 1 − 0.261 exp −0.000777𝑇! ! Option 1: Idso and Jackson (1969)
(16)
1 + 0.17𝐶 ! 9.365×10!! 𝑇! + 273.15 ! Option 2: Swinbank (1963)
𝑒! !/!
𝜀!∗ = 1 + 0.275𝐶 0.642 Option 3: Brutseart (1975)
𝑇!
!
𝑒! !
1 − 𝐶 !.!"# 1.24 + 0.955 𝐶 !.!"# Option 4: Yajima and Yamamoto (2014)
𝑇!
where, C is the cloud cover fraction (0-1), and options 1-4 are chosen via the cloudmode variable. Note that cloud cover is
typically reported in octals (1-8) with each value depicting a fraction of 8. So a value of 1 would correspond to a fraction of
0.125. Some data may also include cloud type and their respective heights. If this is the case, good results have been
reported by averaging the octal values for all kinds of cloud cover to get the total cloud cover average value.
If longwave radiation data does not exist and cloud data is also not available, but solar irradiance is measured, then it is
possible to use get GLM to compare the measured and theoretical (BCSM) solar irradiance in to approximate the cloud
fraction. This option utilises the above relation in Eq 8 to compute 𝜙!" and clouds are approximated by assuming that
𝜙!" !"# 𝜙!" !"#$ = 𝑓(𝐶). Please note that if neither shortwave or longwave radiation are provided, then the model will use
the BCSM to compute incoming solar irradiance and cloud cover will be assumed to be 0.
10
Oct 2014
Sensible and latent heat transfer
The model accounts for the surface fluxes of sensible heat and latent heat using commonly adopted bulk aerodynamic
formulae. For sensible heat:
𝜙! = −𝜌! 𝑐! 𝐶! 𝑈! 𝑇! − 𝑇! (17)
where cp is the specific heat capacity of air (1005 J/kg/°C), CH is the bulk aerodynamic coefficient for sensible hear transfer
(~1.3x10-3), Ta the air temperature (°C) and Ts the temperature of the surface layer (°C). The air density is in kg m-3 and
computed from 𝜌! = 0.348 (1 + 𝑟)/(1 + 1.61𝑟) 𝑝/𝑇! , where r is the mixing ratio, 𝑝 is air pressure in hPa and assuming the
gas constant.
For latent heat:
𝜅
𝜙! = −𝜌! 𝐶! 𝜆 𝑈! 𝑒! 𝑇! − 𝑒! 𝑇! (18)
𝑝
where CE is the bulk aerodynamic coefficient for latent heat transfer, ea the air vapour pressure and es the saturation vapour
pressure (hPa) at the surface layer temperature (°C) and 𝜅 is the ratio of molecular weight of water to molecular weight of
air ( = 0.622). The vapour pressure can be calculated by the following formulae:
!!
𝑒! 𝑇! = 𝑒𝑥𝑝 2.3026 7.5 + 0.7858 Option 1 : TVA (1972) - Magnus-Tetens
!! !!"#.!
!".!"# !!
𝑒! 𝑇! = 𝑒𝑥𝑝 6.1094 Option 2 : August-Roche-Magnus (19)
!! !!"#.!"
!"!!.!"##$ !!
!.!"#$%&!%
𝑒! 𝑇! = 10 !! !!"#.!" Option 3 : Tabata (1973) - Linear
𝑅𝐻
𝑒! 𝑇! = 𝑒 𝑇 (20)
100 ! !
Correction for non-neutral atmospheric stability

For long time integrations (i.e. seasonal), the bulk-transfer coefficients for momentum, CD, sensible heat, CH, and latent
heat, CE, can be assumed approximately constant because of the negative feedback between surface forcing and the
water body’s temperature response (e.g. Strub and Powell, 1987). At finer timescales (hours to weeks), the thermal inertia
of the water body is too great and so the transfer coefficients must be specified as a function of the degree of atmospheric
stratification experienced in the internal boundary layer that develops over the water. Monin and Obukhov (1954)
parameterised the stratification seen in the air column using the now well-known stability parameter, z/L, which is used to
define corrections to the bulk aerodynamic coefficient CH and CE, using the numerical scheme presented in Appendix B.
The corrections may be optionally applied in the model, and requires measurement of windspeed, air temperature and
relative humidity within the internal boundary layer over the lake surface and specification at an hourly resolution.
Still-air limit
The above formulations only apply so long as sufficient wind exists and creates a defined boundary layer over the surface
of the water. As the wind tends to zero (the ‘still-air limit’) equations (16-17) are no longer appropriate as they do not
account for free-convection directly from the water surface. This is a relatively important phenomenon for small dams,
ponds and wetlands since they tend have small fetches that limit the build up of wind speed, and they can have surface
temperatures warmer than the atmosphere for considerable periods, and they are often sheltered from the wind.
Therefore, in some lakes we need to augment Eqs 16-17 with additional calculations to ensure that low wind-speed results
are better captured. The flux estimates can be modified by calculating the evaporative and sensible heat flux values for Ux
= 0 and the given Ux and taking the maximum magnitude of this pair as the result, i.e.,
∗
𝜙!,! = max (𝜙!,! , 𝜙!,!! ) (21)
11
Oct 2014
where 𝜙! is the zero-wind flux, and applies for both evaporative and sensible heat fluxes. 𝜙!,! is calculated from the
equations outlined above and 𝜙! calculations are given below. The two zero-wind speed heat flux equations are taken
from TVA (1972), but modified slightly to return power flux densities in SI units (i.e., Wm-2). The zero wind speed
evaporative mass flux calculation is described as:
𝜙!! = 𝜌! 𝜆 𝛼! 𝐶! − 𝐶! (22)
𝜙!! = 𝛼! 𝑇! − 𝑇!
! !
𝑣 𝜌! − 𝜌!
𝛼! = 2.283×10!! 𝜉 𝑔 (23)
𝑐! 𝜌! 𝜌! 𝜈 𝑎
! !
𝜌! − 𝜌!
𝛼! = 2.283×10!! 𝜉 𝑣 𝑔
𝜌! 𝜈 𝑎
where 𝐶 = 𝜅 𝑒 𝑝, with the appropriate vapour pressure values, e, for both surface and ambient atmospheric values. Here, v
is the molecular heat conductivity of air (0.1 kJ m-1 hr-1 K-1), ν is the kinematic viscosity of the air (0.0548 m2 hr-1], ρo is the
density of the saturated air at the water surface temperature, ρs is the density of the surface water, ξ is a roughness
correction coefficient for the lake surface (0.5), a is the molecular heat diffusivity of air (0.077 m2 hr-1). Note that the impact
of low wind speeds on the drag coefficient is captured by the modified Charnock relation (Eq. A24), which includes an
additional term for the smooth flow transition (see also Figure A1).
Wind-sheltering
Hipsey et al. (2003) presented a simple adjustment to the bulk transfer equation to account for the effect of wind-
sheltering around small dams. The method employs the use of the shelter index which is well suited to one-dimensional
application by accounting for the length scale associated with the vertical obstacle relative to the horizontal length scale
associated with the dam itself (see also Condie and Webster, 2001). A modified form of the shelter index approximation
has been implemented that reduces the effective surface area for heat and momentum fluxes as:
𝐴!
𝐴! = 𝐴! tanh (24)
𝐴!
where AC is the critical area. In GLM, the ratio of the effective area to the total area of the lake 𝐴! 𝐴! is then used to scale
𝑈! as a means of capturing the average wind speed over the entire lake surface.
Snow and ice dynamics

The algorithms for GLM ice and snow dynamics are based on previous ice modelling studies (Patterson and Hamblin,
1988; Gu and Stefan, 1993; Rogers et al., 1995; Vavrus et al., 1996; Launiainen and Cheng, 1998). To solve the heat transfer
equation, the ice model uses a quasi-steady assumption that the time scale for heat conduction through the ice is short
relative to the time scale of meteorological forcing (Patterson and Hamblin, 1988; Rogers et al., 1995).
The steady-state conduction equations, which allocate shortwave radiation into two components, a visible (A1=70%) and
an infra-red (A2=30%) spectral band, which are used with a three-component ice model that includes blue ice (or black
ice), snow ice (or white ice) and snow (see Eq. 1 and Fig. 5 of Rogers et al., 1995). Snow ice is generated in response to
flooding, when the mass of snow that can be supported by the ice cover is exceeded (see Eq. 13 of Rogers et al., 1995). By
assigning appropriate boundary conditions to the interfaces and solving the quasi-steady state of heat transfer
numerically, we determine the upward conductive heat flux between the ice or snow cover and the atmosphere, 𝜙! . The
estimation of 𝜙! involves the application of an empirical equation (Ashton, 1986) to estimate snow conductivity (Ks) from
its density, where the density of snow is determined as outlined in Figure 4.
At the ice (or snow) surface, a heat flux balance is employed to provide the condition for surface melting,
𝜙! 𝑇! + 𝜙!"# 𝑇! = 0 𝑇! < 𝑇!
𝑑ℎ! (25)
= −𝜌𝐿 𝑇! = 𝑇!
𝑑𝑡
12
Oct 2014
where L is the latent heat of fusion (see physical constants, Table 2), hi is the height of the upper snow or ice layer, t is
time, ρ is the density of the snow or ice, determined from the surface medium properties, T0 is the temperature at the solid
surface, Tm is the melt-water temperature (0oC) and φnet(T0) is the net incoming heat flux, at the solid surface:
𝜙!"# 𝑇! = 𝜙!"#$ − 𝜙!"#$% 𝑇! + 𝜙! 𝑇! + 𝜙! 𝑇! + 𝜙! 𝑇! (26)
where φLWin and φLWout are incoming and outgoing longwave radiation, φH and φE are sensible and evaporative heat fluxes
between the solid boundary and the atmosphere, and φR is the heat flux due to rainfall. These heat fluxes are calculated as
above with modification for determination of vapor pressure over ice or snow (Gill, 1982) and the addition of the rainfall
heat flux (Rogers et al., 1995). T0 is determined using a bilinear iteration until surface heat fluxes are balanced (i.e. φ0(T0) = -
φnet (T0)) and T0 is stable (± 0.001oC). In the presence of ice (or snow) cover, surface temperature T0 > Tm indicates that
energy is available for melting. The amount of energy for melting is calculated by setting T0 = Tm to determine the
reduced thickness of snow or ice (as shown in Eq 25).
Accretion or ablation of ice is determined through the heat flux at the ice-water interface, qf,. Solving for heat conduction
through ice yields:
𝑞! = 𝑞! − 𝐴! 𝜙!" 1 − exp −𝐾!! ℎ!"#$ − 𝐾!! ℎ!!!"# − 𝐾!! ℎ!"#$ − 𝐴! 𝜙!" 1 − exp −𝐾!! ℎ!"#$ − 𝐾ℎ!!!"# − 𝐾!! ℎ!"#$ − 𝑄!!!"# ℎ!"#$ (27)
where 𝜙!" is the shortwave radiation penetrating the surface, K refers to the light attenuation coefficient of the ice and
snow components designated with subscripts s, w and e for snow, blue ice and snow ice respectively, and h refers to the
thickness of snow, white (snow ice) and blue ice. Qwhite is a volumetric heat flux for formation of snow ice, which is given in
Eq. 14 of Rogers et al. (1995). Ice and snow light attenuation coefficients in GLM are fixed to the same values as those
given by Rogers et al. (1995). Reflection of shortwave radiation from the ice or snow surface is a function of surface
temperature and ice and snow thickness (see Table 2, Vavrus et al., 1996); values of albedo derived from these functions
vary from 0.08 to 0.6 for ice and from 0.08 to 0.7 for snow.
The imbalance between qf and the heat flux from the water to the ice, qw, gives the rate of change of ice thickness at the
interface with water:
𝑑ℎ!"#$ 𝑞! − 𝑞!
= (28)
𝑑𝑡 𝜌!"#$ 𝐿
where ρblue is the density of blue ice and qw is given by a finite difference approximation of the conductive heat flux from
water to ice:
∆𝑇
𝑞! = −𝐾! , (29)
∆𝑧
where Km is molecular conductivity and ΔT is the temperature difference between the surface water and the bottom of the
ice, which occurs across an assigned depth Δz. A value for Δ z of 0.5 m is usual, based on the reasoning given in Rogers et
al. (1995) and the typical vertical resolution of a model simulation (0.125 – 1.5 m). Note that a wide variation in techniques
and values is used to determine the basal heat flux immediately beneath the ice pack (e.g., Harvey, 1990).
Figure 4 shows the overall algorithm approach to update ice cover, snow cover and water depth. The ice cover equations
are applied when water temperature first drops below 0 °C. The ice thickness is set to its minimum value of 0.05 m, which
is suggested by Patterson and Hamblin (1988) and Vavrus et al. (1996). The need for a minimum ice thickness relates
primarily to horizontal variability of ice cover during the formation and closure periods. The ice cover equations are
discontinued and open water conditions are restored in the model when the thermodynamic balance first produces ice
thickness < 0.05 m. The effects of snowfall, rainfall, and compaction of snow are described through appropriate choice of
one of several options, depending on the air temperature and whether ice or snow is the upper boundary (Figure 4).
Density of fresh snowfall is determined as the ratio of measured snowfall height to water-equivalent height, with any values
exceeding the assigned maximum snow density (ρmax = 300 kg m-3) truncated to the upper limit. The snow compaction
model is based on the exponential decay formula of McKay (1968), with selection of snow compaction parameters based
on air temperature (Rogers et al., 1995) as well as on rainfall or snowfall. The approach of snow compaction used by Rogers
et al. (1995) is to set the residual snow density to its maximum value when there is fresh snowfall. This method is found to
produce increases in snow density that are too rapid when there is only light snowfall. As a result a gradual approach to
increasing snow compaction is adopted.
13
Oct 2014
Figure 4: Decision tree to update ice cover, snow cover and water depth according to snow compaction, rainfall
(P) and snowfall (S) on each day, and depth of snow cover (hsi) and snow density ( ρ s i ) for the previous day. Refer to
Table 1 for definitions of other variables.
Stratification and vertical mixing
Surface Mixed Layer:

GLM works on the premise that the balance between the available energy, ETKE, and the required energy to undergo
mixing, EPE, provides for the surface mixed layer (SML) deepening rate dzSML/dt. For an over view of the dynamics readers
are referred to early works on bulk mixed layer depth models by Kraus and Turner (1967) and Kim (1976), later more fully
extended by Imberger & Patterson (1981) which is the basis for the model design. In this model, the available kinetic
energy is calculated due to contributions from wind stirring, shear production between layers, convective overturn, and
Kelvin-Helmholtz (K-H) billowing.
They may be combined and summarised for ETKE as (Hamilton and Schladow, 1997):
𝑢!! 𝑑𝜉 𝑢! 𝜉 𝑑𝑢!
𝐸!"# = 0.5𝐶! 𝑤∗! Δt + 0.5𝐶! 𝐶! 𝑢∗! Δt + 0.5 𝐶! 𝑢!! + + Δ𝑧!!! (30)
!"#$%!&'$% !"#$%&$'
6 𝑑z!"# 3 𝑑z!"#
!"#$ !"#$$#%&
!!!"# !"#$%&'(#)
!!! !"#$%&'(#)
where and 𝜉 is the K-H billow length scale (described below), 𝑢! is the shear velocity at the interface of the mixed layer,
and 𝐶! , 𝐶! , and 𝐶! are mixing efficiency constants. The energy required to lift up water at the bottom of the mixed layer,
denoted here as the layer 𝑘 − 1, with thickness ∆ℎ!!! , and accelerate it to the mixed layer velocity is required for mixing to
occur. This also accounts for energy consumption associated with K-H production and expressed as, EPE:
14
Oct 2014
Δ𝜌 𝑔𝜉 ! 𝑑 Δ𝜌 𝑔𝜉Δ𝜌 𝑑𝜉 (31)
𝐸!" = 0.5𝐶! 𝑤∗! + 𝐶! 𝑢∗! ! !
+ 𝑔 z!"# + + Δ𝑧!!!
𝜌! 24𝜌! 𝑑z!"# 12𝜌! 𝑑z!"#
!""#$#%!&'()
!"#$"%& !!! !"#$%&'()"#
where, z!"# is the thickness of the surface mixed layer. To numerically resolve the above equations we must sequentially
compute the different components of the above expressions in light of the layer structure, and GLM follows the algorithm
in Imberger and Patterson (1981) whereby we first undertake cooling and combination of layers due to convection, then
undertake stirring and then computing shear and K-H mixing.
To compute mixing due to convective cooling we compute the value for 𝑤∗ , which is the turbulent velocity scale
associated with convection. The model adopts the algorithm used in Imberger and Patterson (1981; Eq 32), whereby the
potential energy that is released by mixed layer deepening is computed by looking at the moments of the different layers
in the surface mixed layer (from layer K to NLEV):
!!"# !!"#
𝑔 (32)
𝑤∗! = 𝜌! Δz! h! − h!"# 𝜌! Δz!
𝜌!"# Δt
!!! !!!
where 𝜌!"# is the mean density of the mixed layer including the combined layer, 𝜌! is the density of the kth layer, Δz! is the
height difference between two consecutive layers within the loop (Δz! = h! − h!!! ), h! is the mean height of layers to be
mixed ( h! = 0.5[ h! + h!!! ] ), and h!"# is the epilimnion (surface mixed layer) mid height, calculated from: h!"# =
0.5 ℎ!"#$ + h!!! .
The velocity scale 𝑢∗ is associated with wind stress and calculated according to the wind strength:
𝑢∗ ! = 𝐶! 𝑈!! (33)
where 𝐶! is the drag coefficient for momentum. The model first computes the following check to see if the stirring energy
is enough to overcome the energy required to mix the k-‐1 layer, i.e., mixing occurs if:
𝐶! 𝑤∗! + 𝐶! 𝑢∗! Δt ≥ 𝑔!! z!"# + 𝐶! 𝑤∗! + 𝐶! 𝑢∗! ! !

Δ𝑧!!! (34)
!!
and 𝑔!! = is the reduced gravity between the mixed layer and k-‐1 layer. If the condition is not met the energy is stored
!!
for the next time-step.
Once stirring is completed mixing due to velocity shear is applied. Velocity shear at the interface is approximated from:
𝑢∗ ! 𝑡 (35)
𝑢! = + 𝑢!
𝑧!"#
where t is a time value over which it has been operating, considered relative to tshear which is the time beyond which shear
production doesn't occur (ie., 𝑢! = 0 if the time exceeds tshear). This cut off time assumes use of only the energy produced
by shear at the interface during the half seiche period, Ti, and modified to account for damping:
!!
𝑇! (36)
𝑡!!!"# = 𝑇! 1 + 0.59 1 − 𝑐𝑜𝑠ℎ −1
𝑇!
where 𝑇! is the time-scale of damping (see Spigel, 1978). The wave period is approximated based on the stratification as
𝑇! = 𝐿!!"# 2𝑐, where 𝐿!"#$ is the length of the domain at the thermocline and c is the internal wave speed. Once the
velocity is computed, the energy for mixing from velocity shear is compared to that required for lifting and accelerating
the next layer down and layers are combined:
𝑢!! 𝑧!"# + ∆𝜉 𝑢! 𝜉∆𝑢! 𝜉Δ𝑧!!! ∆𝜉 (37)

0.5 𝐶! + + 𝑔!! 𝜉 − ≥ 𝑔!! z!"# + 𝐶! 𝑤∗! + 𝐶! 𝑢∗! ! !
Δ𝑧!!!
6 3 24𝑧!"# 12
where the K-H length scale is 𝜉 = 𝐶!" 𝑢!! 𝑔!"

! !
, and ∆𝜉 = 2 𝐶!" 𝑢! ∆𝑢! 𝑔!" ; in this case the reduced gravity is computed
from the difference between the epilimnion and hypolimnion, and 𝐶!" is a measure of the billow mixing efficiency.
15
Oct 2014
Once shear mixing is done, the model checks the resultant density interface to see if it is unstable to shear (ie. K-H billows
would be expected to form). This occurs if the gradient is less that the K-H length scale, and then if K-H mixing is required
layers are further split and subject to mixing using an algorithm similar to above.
Deep Mixing:
Mixing below the SML in lakes, in the deeper stratified regions of the water column, is modelled using a characteristic
vertical diffusivity, 𝐾! = 𝐾! + 𝐾! , where 𝐾! is the fixed molecular diffusivity of scalars. The model adopted in GLM is based
on the derivation by Weinstock (1981) that is described as being suitable for regions displaying weak or strong
stratification, whereby diffusivity increases with dissipation and decreases with heightened stratification:
𝛼 !"# 𝜀!"#
𝐾! = (38)
𝑁 ! + 0.6 𝑘 !"# ! 𝑢∗ !
where 𝛼 !"# is the mixing efficiency of hypolimnetic TKE (~0.8 in Weinstock, 1981) and 𝑘 !"# is the turbulence wavenumber:
12.4 𝐴!"#
𝑘 !"# = (39)
𝑉 ∆𝑧!"# 10!
!
and 𝑢∗ = 1.612 ×10!! 𝑈! . The term 𝑁 ! is the Brunt–Väisälä (buoyancy) frequency defined as:
𝑔∆𝜌 𝑔 𝜌!!! − 𝜌!!! (40)

𝑁! = ≈
𝜌∆𝑧 𝜌!"# ℎ!!! − ℎ!!!
Estimating the turbulent dissipation rate can be complex but GLM adopts the simple approach as described in Fischer et
al. (1980) where a “net dissipation” is approximated by assuming dissipation is in equilibrium with energy inputs from
external drivers:
𝜀!"# ≈ 𝜀!"# = 𝐸!"#$ + 𝐸!"#$%& (41)
which is expanded and calculated per unit volume as:
!!"#
1 𝑚 𝐶! 𝜌! 𝑓! 𝑈! ! 𝐴! 1 (42)
𝜀!"# = + 𝑔 ∆𝜌! 𝑄! ℎ!"# − ℎ!
𝑉𝜌 10! 10! 𝑉!"# 𝜌 10!
!
!"#$ !" !"#$%&' !" !"#$ !"#$ !" !"#$%&' !"#$ !" !"#$%&'
The diffusivity is calculated according to Eq 42, but since the dissipation is assumed to concentrate close to the level of
strongest stratification, the “mean” diffusivity is modified to decay exponentially with distance from the thermocline:
0 ℎ! ≥ ℎ!"# − 𝑧!"# (43)

𝐾!! = −(ℎ!"# − 𝑧!"# − ℎ! )!
𝐾! 𝑒𝑥𝑝 ℎ! < ℎ!"# − 𝑧!"#
𝜎
where is the 𝜎 variance the N2 distribution below ℎ!"# and scales the depth over which mixing decays.
Once the diffusivity is approximated, the diffusion of any scalar, 𝐶, between two layers is numerically accounted for by the
following mass transfer expressions:
exp (−𝑓)∆𝑧! ∆𝐶
𝐶!!! = 𝐶 + (44)
(∆𝑧!!! + ∆𝑧! )
exp (−𝑓)∆𝑧!!! ∆𝐶
𝐶! = 𝐶 −
(∆𝑧!!! + ∆𝑧! )
where 𝐶 is the weighted mean concentration of 𝐶 for the two layers, and ∆𝐶 is the concentration difference between
them. 𝑓 is related to the diffusivity according to:
𝐾!!!! + 𝐾!!
𝑓= ∆𝑡 (45)
∆𝑧!!! + ∆𝑧! !
The above diffusion algorithm is run once up the water column and once down the water column as a simple explicit
method for capturing diffusion to both the upper and lower layers.
16
Oct 2014
Inflows and outflows
Inflows can be specified as surface runoff from the surrounding lake domain (described above, Eq 5), rivers entering at the
surface of the lake or submerged inflows. Any number of inflows to the lake body can be specified and these are applied
at the end of the sub-daily loop, i.e. once a day.
Three forms of outflows are included in GLM, ground water seepage, outflow from a specified depth or overflow.
River inflows
For river inflows, depending on the density of the river water, the inflow will form a positive or negatively buoyant
intrusion. As the inflow crosses layers it will entrain water out of them, until it reaches a level of neutral buoyancy and
undergoes insertion. Therefore, when it reaches its point of neutral buoyancy a new layer of thickness dependent on the
inflow volume at that time (including additions from entrainment) is created. Following insertion, the inflow layer may then
amalgamate with adjacent layers depending on numerical criteria within the model for combining or splitting layers.
The rate of entrainment of the intrusion, 𝐸, can be calculated in a number of ways. For simplicity in GLM, the rate has
been adapted from the first approximation in Fisher et al. (1979):
𝐶!! !/! (46)

𝐸 = 1.6
𝑅𝑖
where 𝐶!! is the user specified drag coefficient for the inflow. The Richardson’s number is adapted from Fisher et al. (1979)
as:
𝐶!! 1 + 0.21 𝐶!! sin 𝛼!"# (47)

𝑅𝑖 =
sin 𝛼!"# tan 𝜙!"#
where 𝛼!"# is the stream half angle and 𝜙!"! is the slope of the inflow at the point where it meets the water body
(Figure 5).
Figure 5: Schematic showing inflow insertion, entrainment and slope, 𝝓𝒊𝒏𝒇 and half angle, 𝜶𝒊𝒏𝒇 of inflowing rivers.
As the inflow parcel travels through the layers, the increase in inflow thickness due to entrainment is estimated as:
ℎ! = 1.2𝐸𝑑𝑥 + ℎ!!! (48)

where ℎ! is the inflow thickness, 𝐸 is the entrainment rate and 𝑑𝑥 is the distance travelled by the inflowing water,
calculated from the flow rate and inflow thickness. The initial estimation of the intrusion height is computed from
(Imberger and Patterson, 1981; Antenucci et al., 2005):
17
Oct 2014
!/!
𝑅𝑖 (49)
ℎ! = 2𝑄!"# ! 𝑡𝑎𝑛! 𝜙!"#
𝑔′!"# ′
where 𝑄!"# is the inflow rate provided as a boundary condition and 𝑔′ is the reduced gravity of the inflow given as:
𝜌!"# − 𝜌! (50)
𝑔′!"# = 𝑔
𝜌!
where 𝜌!"# is the density of the inflow and 𝜌! the density of the surface layer. The distance travelled by the inflow aliquot,
𝑑𝑥, is estimated as the distance travelled in the vertical and the slope of the inflow river, 𝜙!"# and given by:
𝑑𝑧
𝑑𝑥 = (51)
sin 𝜙!"#
where 𝑑𝑧 is the distance travelled in the vertical. The velocity of the inflow aliquot for that day is then calculated as:
𝑄!"#
𝑢 = ℎ!! (52)
tan 𝛼
Following conservation of mass, the flow is estimated to increase according to (Imberger and Patterson, 1981; Antenucci
et al., 2005):
!/!
ℎ! (53)
𝑄! = 𝑄!!! −1
ℎ!!!
The above entrainment and insertion algorithm is repeated fro each inflow. Aside from importing mass into the lake, river
inflows also contribute turbulent kinetic energy to the hypolimnion as discussed in the Deep Mixing section above (e.g.,
Eq 42).
Submerged inflows
Submerged inflows are inserted at the specified depth with zero entrainment. The submerged inflow layer is then mixed
with adjacent layers above or below depending on the density difference until neutral buoyancy is reached.
Withdrawals
Outflows can be specified at any depth over the water column and will draw water from the adjacent layer, layers above or
below depending on the strength of discharge and stability of the water column according to the following algorithms.
The three types of outflow, seepage, withdraw and overflow all use the same algorithms with overflow volume calculated
by the volume of water in excess of maximum storage once rainfall, evaporation and all inflows and outflows have been
accounted for.
The thickness of the withdrawal layer is dependent on the calculation of the internal Froude (𝐹𝑟) and Grashof (𝐺𝑟) numbers
and a parameter, R (Fisher et. L 1979):
𝑄!"#$
𝐹𝑟 = (54)
!
𝑁!"#$ 𝑊!"#$ 𝐿!!"#$
!
𝑁!"#$ 𝐴!! (55)
𝐺𝑟 = !
𝑣!"#$
!
𝑅 = 𝐹𝑟𝐺𝑟 ! (56)
!
Where 𝑊!"#$ , 𝐿!"!" and 𝐴! are the width, length and area of the lake at the outlet elevation, and 𝑣!"#$ is the vertical
diffusivity of momentum averaged over the withdrawal layer and the Brunt- Väisälä frequency averaged over the thickness
!
of the withdrawal layer, 𝑁!"#$ is calculated as:
! 𝑔 𝜌!"#$ − 𝜌!
𝑁!"#$ = (57)
𝑑𝑧 𝜌!"#$
where 𝑑𝑧 is the thickness of the withdrawal layer, 𝜌!"#$ is the density of the lake at the height of withdrawal and 𝜌! is the
density of the lake at the edge of the withdrawal layer.
18
Oct 2014
The thickness of the withdrawal layer is then calculated as follows (Fisher et al. 1978):
𝛿!"#$ = 2𝐿!"#$ 𝐺𝑟 !! !
(58)
𝛿!"#$ = 2𝐿!"#$ 𝐹𝑟 ! !
The proportion of fluid withdrawn from each layer either above or below the layer of the outlet elevation is determined
using a curve that fits the region of fluid drawn in a given time.
To calculate the width and length of the lake at the height of outflow the following assumptions are made:
1. That the lake shape approximates as an ellipse.

2. The ratio of length to width at height of outflow is the same as that at the lake crest.
The length of the lake at the outflow height, 𝐿!"#$ and the lake width, 𝑊!"#$ are given by:
4 𝐿!"#$% (59)
𝐿!"#$ = 𝐴!"#$
𝜋 𝑊!"#$%
𝑊!"#$%
𝑊!"#$ = 𝐿!"#$
𝐿!"#$%
where 𝐴!"#$ is the area of the lake at the outflow height, 𝐿!"#$% is the length and 𝑊!"#$% the width of the lake at the crest
height.
Seepage
Seepage of water from the bottom layer is also configurable within the model, for example, as might be required in a
wetland simulation. Seepage is configured to leave the lake at a constant rate:
𝑑ℎ!
= −𝐺 (60)
𝑑𝑡
where hB is the depth of the bottom-most layer at any time, and G is the seepage rate (m day-1). 𝐺 is constrained within the
model to ensure no more than 50% of the layer can be reduced in any one time-step. Note that in shallow simulations, a
single layer may form, in which case the surface and bottom later are the same and Eq 4 and 60 are combined.
Bottom stress
Wind-induced resuspension of sediment from the bed of shallow lakes is sporadic and occurs as the waves created at the
water surface create oscillatory currents that propagate down to the lake-bed. GLM does not predict resuspension and
sediment concentration directly, but computes the bottom shear stress for later use by sediment and water quality
modules that are within FABM-AED. Nonetheless, even without this sophistication the model can identify the areal extent
and potential for bed-sediment resuspension by computing the area of the lake over which the bed shear stress exceeds
some critical value required for resuspension to occur.
To compute the stress at the lake bottom we estimate the surface wave conditions using a simple, fetch-based, steady
state wave model (Laenen and LeTourneau, 1996; Ji 2008). The wave geometry (wave period, significant wave height and
wave length), is predicted based on the windspeed and fetch over which the waves develop (Figure 6), calculated as:
𝐴! (61)
𝐹=2 𝜋
Using this model, the wave period, T, is calculated from fetch as:
!" !.!!!
𝑈! 0.0379 ! (62)
!!
𝑇 = 7.54 𝑡𝑎𝑛ℎ 𝜉 𝑡𝑎𝑛ℎ
𝑔 𝑡𝑎𝑛ℎ 𝜉
where:
19
Oct 2014
!.!"#
𝑔𝑑!"# (63)
𝜉 = 0.833
𝑈!!
and ℎ!"# is the average lake depth. Wave length is then estimated from:
𝑔𝑇 ! 2𝜋 𝑑!"# (64)
𝐿= 𝑡𝑎𝑛ℎ !! !
2𝜋
!!
and wave height from:
!" !.!
𝑈!! 0.00565 ! (65)
!!
𝐻! = 0.283 𝑡𝑎𝑛ℎ 𝜁 𝑡𝑎𝑛ℎ
𝑔 𝑡𝑎𝑛ℎ 𝜁
where
!.!"
𝑔𝑑!"# (66)
𝜁 = 0.53
𝑈!!
Based on these properties the orbital wave velocity at depth (in the i th layer) is calculated as:
𝜋𝐻!
𝑈!"#! = (67)
!! !!
𝑇𝑠𝑖𝑛ℎ
!
The total shear stress at the lake bed is calculated as:
1
𝜏! = 𝜌! 𝑓! 𝑈!"#! ! + 𝑓! 𝑈!! ! (68)
2
where 𝑈! is the mean velocity of the layer, computed during the mixing calculations (Eq 33). The friction factors use D (a
typical particle diameter). For the current stress we compute (𝑓! = 0.24 log 12𝑑!"# 2.5𝐷 and for waves, based on:
! !!.!"
𝑓! = exp −5.977 + 5.213 Option 1 : Laenen and LeTourneau, 1996 (69)
!.!!
!!"# ! !!.!"
𝑓! = 0.00251 exp 5.213 Option 3 : Kleinhans & Grasmeijer (2006)
!!"
!!"!! !
𝑓! = Option 3 : Le Roux (2007)
! ! !!
Figure 6: Slope, 𝝓𝒊𝒏𝒇 and half angle, 𝜶𝒊𝒏𝒇 of inflowing rivers.
20
Oct 2014
Table 1. Summary of GLM physical parameters with recommended values and references.
Symbol glm.nml ID Description Units Default Reference Comments
Model Structure
Standardised for multi-
ℎ!"# min_layer_thick Minimum layer thickness m 0.5 -
lake comparison
Should be estimated
ℎ!"# max_layer_thick Maximum layer thickness m 1.5 -
relative to lake depth.
Lake Properties
Should be measured, e.g.
Extinction coefficient for PAR mean of simulation
𝐾! Kw
radiation
m-1 0.2 Lake specific
period. Can be estimated
from Secchi depth.
Xenopoulos
Critical area below which wind
𝐴! critical_area
sheltering may occur
m2 107 and Schindler
(2001)
Surface Thermodynamics
Bulk aerodynamic coefficient for Fischer et al. From Hicks' (1972)
𝐶! ch
sensible heat transfer
- 0.0013
(1979) collation of ocean and
Bulk aerodynamic coefficient for Fischer et al. lake data; many studies
𝐶! ce
latent heat transfer
- 0.0013
(1979) since use similar values.
Internally calculated if
Bulk aerodynamic coefficient for Fischer et al. atmos stability correction
𝐶! cd
transfer of momentum
- 0.0013
(1979) is on.
𝜆 - Latent heat of evaporation J kg-1 2.453x106 Standard Not adjustable in glm.nml

Water only, no ice
𝜀! - Emissivity of the water surface - 0.985 Standard
Ice or snow
𝜎 - Stefan-Boltzmann constant Wm K-2 -4

5.67x10 -8
Not adjustable in glm.nml
Mixing Parameters
Yeates & Selected by Yeates et al
Mixing efficiency - convective
𝐶! coef_mix_conv
overturn
- 0.2 Imberger (2004) from a range given
(2003) in Spigel et al. (1986)
Spigel et al.
𝐶! coef_wind_stir Mixing efficiency - wind stirring - 0.23
(1986)
From Wu 1973
Mixing efficiency - shear Sherman et al. Best fit of experiments

𝐶! coef_mix_shear
production
- 0.3
(1978) reviewed
Mixing efficiency - unsteady
𝐶! coef_mix_turb
turbulence (acceleration)
- 0.51
Mixing efficiency - Kelvin- Sherman et al.

𝐶!" coef_mix_KH
Helmholtz turbulent billows
- 0.3
(1978)
"a good rule of thumb..."
Mixing efficiency of hypolimnetic Weinstock General diffusivities in

𝐶!"# coef_mix_hyp
turbulence
- 0.5
1981 Jellison and Melack (1993)
Inflows & Outflows

Set based on inflow
𝐶! strmbd_drag streambed_drag - 0.016
stream type
𝐺 seepage_rate Seepage rate m day-1 0 Site specific
Snow & Ice

Waveband 1, snow ice light
𝐾!! - m-1 48.0
extinction
Waveband 2, snow ice light
𝐾!! - m-1 20.0
extinction
Waveband 1, blue ice light
𝐾!! - m-1 1.5
extinction
Waveband 2, blue ice light
𝐾!! - m-1 20.0
extinction
Waveband 1, snow light
𝐾!! - m-1 6
extinction
Waveband 2, snow light
𝐾!! - m-1 20
extinction
Distance of hear transfer, ice
𝐷! - m 0.039
water
21
Oct 2014
Symbol glm.nml ID Description Units Default Reference Comments
𝜌!!!"# - Density, snow ice kg m-3 890
𝜌!"#$ - Density, blue ice kg m-3 917
𝜌!"#$ - Density, snow kg m-3 Variable
𝑐!" - Heat capacity, ice kJ kg-1 oC-1 2.1
𝑐!" - Heat capacity, ice kJ kg-1 oC-1 4.2
𝐾! - Compaction coefficient - Variable
𝐾! - Thermal conductivity, snow ice W m-1 oC-1 2.0
𝐾! - Thermal conductivity, blue ice W m-1 oC-1 2.3
𝐾! - Thermal conductivity, snow W m-1 oC-1 Variable
𝐾! - Thermal conductivity, sediment W m-1 oC-1 1.2
𝐾! - Thermal conductivity, water W m-1 oC-1 0.57
𝐿 - Latent heat of fusion kJ kg-1 0334
Bottom Stress
𝐷 - Typical particle diameter m
22
Oct 2014
Setup & Operation
Overview
This section gives a description of the structure of a GLM setup is described. GLM requires a configuration files and
several time-series input files and integrates with FABM-AED or AED2 for water quality simulations (Figure 7).
Configuration •  GLM2.nml!
& Parameters •  FABM.nml!
FABM
Specialist AED module
Parameter files:
AED aed_phyto_pars.nml
• InflowX.csv!
• InflowY.csv!
GLM.exe aed_zoop_pars.nml
aed_geochem_pars.nml
Time Series • Outflows.csv!
Inputs • Met.csv!
NETCDF file : output.nc!
Sepcific depth time-series

file : depth_xx.csv!
Lake system daily

summary : lake.csv!
Figure 7: Flow diagram showing the files required for operation of the model.
Input files
Physical m odel configuration: glm.nml

The file glm.nml is the main configuration file for the physical model, and some details related to the FABM coupling.
The nml file includes detailed description of the different namelist options for each block; if these value are not present
default values will be assumed. It is a namelist file with blocks for:
• &glm_setup: General simulation info and mixing parameters

• &wq_setup: Details about the coupling with the water quality model (eg. FABM or AED2)
• &time: Time controls
• &morphometry: Lake morphometric information
• &output: Specification of output file details (depths, output frequency and variables to write)
• &init_profiles: Setting initial conditions (depth profiles) of GLM and WQ variables
• &meteorology: Information about surface forcing and meteorology data
• &inflows: Information about inflows
• &outflows: Information about outflows
• &bird: Optional block to input parameters for the Bird solar radiation model
Refer to the example glm.nml files for detailed over view of the layout and required information. The key elements are
described below.
23
Oct 2014
M eteorological configuration and met.csv
A range of options exist for customising the meteorological forcing of the lake simulation.
Key configuration variables users may provide are:
• met_sw: Switch to enable (.true.) or disable (.false.) meteorological forcing.

• snow_sw: Switch to enable (.true.) or disable (.false.) the snow/ice model.
• rain_sw: Switch to enable (.true.) or disable (.false.) rainfall nutrient composition.
• rad_mode: Switch to configure the shortwave/longwave radiation sub-model (Table 2).
• cloud_mode: Switch to configure the atmospheric longwave emissivity sub-model (Eq 16).
• albedo_mode: Switch to configure the shortwave albedo algorithm (Eq 9).
• atm_stab: Switch to enable (.true.) or disable (.false.) non-neutral atmospheric stability.
Details of the meteorological boundary condition data are summarised according to:
• subdaily: Determines whether the model expects to read in sub-daily meteorological data.
• meteo_file: Name of the csv file containing meteorological data.
• time_fmt: Format of the time/data column in the meteorological input file.
• wind_factor: Scaling factor that is used to multiply the wind speed data that is read in.
• rain_factor: Scaling factor that is used to multiply the rainfall data that is read in.
• at_factor: Scaling factor that is used to multiply the air temperature data that is read in.
• rh_factor: Scaling factor that is used to multiply the relative humidity data that is read in.
• sw_factor: Scaling factor that is used to multiply the shortwave data that is read in.
• lw_factor: Scaling factor that is used to multiply the longwave data that is read in.
Details of the parameters used in the model include:
• ce: Bulk-transfer coefficient for latent heat flux calculation under neutral conditions.
• ch: Bulk-transfer coefficient for sensible heat flux calculation under neutral conditions.
• cd: Bulk-transfer coefficient for momentum flux calculation under neutral conditions.
Table 2: Summary of the ways GLM can treat solar radiation and cloud data, as configured through rad_mode.
Solar data creation

rad_mode Data required in met.csv
and treatment of longwave or cloud data
0 Solar (daily) Clouds (daily) Daily solar radiation data is subject to a sine wave
disaggregation to get a sub-daily light time-series. The daily
total energy input equals the daily value provided by the user.
Cloud cover data is used for prediction of longwave radiation.
1 Solar (sub-daily) Clouds (sub-daily) Sub-daily solar radiation data is used directly.
Cloud cover data is used for prediction of longwave radiation.
2 Solar (sub-daily) No Cloud data, Sub-daily solar radiation data is used directly.
LongWave provided Clouds are not used in the model, longwave data is expected
in met.csv and is used directly in the heat balance.
3 Solar (sub-daily) No Cloud data, No Sub-daily solar radiation data is used directly.
Longwave provided No longwave or cloud data is provide to the model, so GLM
will attempt to estimate cloud fraction: 𝜙!" 𝜙!" = 𝑓(𝐶)
4 None Clouds (sub-daily) Sub-daily solar radiation data is computed using the BCSM
model, adjusted for cloud cover: 𝜙!" = 𝜙!" 𝑓(𝐶)
Cloud cover data is used for prediction of longwave radiation
5 None None Sub-daily solar radiation data is computed using the BCSM
model, and clear-sky values will be assumed.
Longwave radiation is predicted with a cloudiness of 0
assumed.
24
Oct 2014
Users must be sure to provide the correct combination of “ShortWave” and “LongWave” or “Clouds” according to their
chosen rad_mode (Table 4).
The meteorological conditions are provided as a time-series of data with a fixed number of columns, as outlined in
Table 3. This maybe included as daily data (subdaily =.false.) or at the time-step of the model simulation (eg.,
hourly) if subdaily =.true. It contains seven compulsory columns, and several optional columns after these
depending on the user-defined configuration switches for snow_sw and rain_sw in the glm.nml file.
Table 3: Flow diagram showing the files required for operation of the model.
met.csv column Units Description
(1) TIME YYYY-MM-DD Date, specified in the format of time_fmt , defaults as:
YYYY-MM-DD HH:MM:SS
time
(2) SHORTWAVE W/m2 Daily average shortwave radiation.

RADIATION
Required if rad_mode = 0, 1, 2 or 3 Note that the daily value is internally distributed to a sub-
ShortWave daily time step by assuming and idealized diurnal cycle.
(3) LONGWAVE W/m2 Longwave radiation input is assumed to be direct

RADIATION incident intensity,.
Required if rad_mode = 2
LongWave
(3) CLOUD COVER - Incoming longwave flux, is estimated from cloud cover
fraction data
Clouds Required if rad_mode = 0, 1 or 4
(4) AIR TEMPERATURE °C Daily average air temperature 10m above the water
surface
AirTemp
(5) RELATIVE HUMIDITY % Daily average relative humidity (0-100%) 10m above the
water surface.
RelHum
(6) WIND SPEED m/s Daily average wind speed 10m above the water surface
WindSpeed
(7) RAINFALL m/day Daily rainfall depth
Rain
(8) SNOWFALL (optional) m/day Daily snowfall depth

(optional – include if snow_sw is T)
Snow
(9-14) RAINFALL WQ mg/L Assumed concentration of WQ variables in the rainfall

DEPOSITION
CONCENTRATIONS (optional – include if rain_sw is T)
(optional)
25
Oct 2014
Configuration of inflows and setup of inflows.csv
A range of options exist for customising the inflow forcing of the lake simulation, in the &inflow section. Any number of
inflows can be simulated by the model with the configuration and filenames set in the glm.nml file.
Several configuration variables and parameters must be provided:
• num_inflows: Number of streams to be simulated. Set to 0 if no streams are included.

• names_of_strms: Names of the inflowing streams/rivers, separated by commas.
• strm_hf_angle: Steambed half-angle for each inflowing stream (Figure 5), separated by commas.
• strmbd_slope: Steambed slope for each inflowing stream (Figure 5), separated by commas.
• strmbd_drag: Steambed drag for each inflowing stream (Table 1), separated by commas.
• submerged: Flag for each inflow indicating if it is a submerged input, separated by commas.
Details of the inflow boundary condition data are summarised according to:
• inflow_fl: Name of the csv file containing the inflow data.

• time_fmt: Format of the time/data column in the inflow input file.
• inflow_factor: Scaling factor that is used to multiply the inflow data that is read in.
• inflow_varnum: Number of variables to be read in for each inflow.
• inflow_vars: List and order of variables being read in. Order must be as in the input file.
For each inflow there is an associated inflow file of the format outlined in Table 4. At this stage the file only accepts daily
data as the inflow calculation is done once a day. It contains four mandatory columns for time, flow, temperature and
salinity, and optional columns for water quality constituents.
inflow.csv column Units Description
(1) TIME YYYY-MM-DD Date, specified in the format of time_fmt , defaults as:
YYYY-MM-DD HH:MM:SS
time
(2) INFLOW ML/day Daily flow rate.

Convert from m3/s by multiplying by 86.4.
flow
(3) STREAMFLOW °C Average daily streamflow temperature

TEMPERATURE
temp
(4) STREAMFLOW mg/L Average daily streamflow salinity

SALINITY
salt
(5 … nWQ+4) mmol/m3 Average daily streamflow water quality constituent

STREAMFLOW WATER concentrations.
QUALITY PARAMETER
CONCENTRATIONS
aed_oxygen_oxy
26
Oct 2014
Configuration of outflows and setup of outflows.csv
Any number of outflow fluxes can be configured and these are set as consecutive columns in the file outflows.csv
(Table 5). Only daily flow rates are required and water quality variables are not required. An additional seepage rate
variable may also be specified, and these details are listed in the &outflow section.
• num_outlet: Number of streams to be simulated. Set to 0 if no streams are included.

• flt_off_sw: Names of the inflowing streams/rivers, separated by commas.
• outl_elevs: Steambed half-angle for each inflowing stream (Figure 4), separated by commas.
• bsn_len_outl: Steambed slope for each inflowing stream (Figure 4), separated by commas.
• bsn_wid_outl: Steambed drag for each inflowing stream (Table 1), separated by commas.
• seepage_rate: In addition to the above outflows, a constant seepage rate (Table 1), can be set.
Details of the inflow boundary condition data are summarised according to:
• outflow_fl: Name of the csv file containing the inflow data.

• time_fmt: Format of the time/data column in the inflow input file.
• outflow_factor: Scaling factor that is used to multiply the inflow data that is read in.
outflow.csv Units Description

column
(1) TIME YYYY-MM-DD Date
time
(2 … nOUT+1) ML/day Daily outflow rates of each outflow

OUTFLOW
flow
Configuring the m odel to run water quality

In GLM v2 and above water quality can be simulated through coupling with FABM, or directly to the AED2 library (Hipsey,
2014). The water quality aspects of the simulation are engaged by including the &wq_setup information in glm.nml.
• wq_lib: Name of the WQ library to be engaged: Either choose ‘aed2’ or ‘fabm’.

• ode_method: Numerical method of solution of the biogeochemical model ODE equations.
• split_factor: Factor used for solution of the FABM biogeochemical ODEs.
• bioshade_feedback: Determines whether the extinction coefficient is updated based on WQ variables.
• repair_state: Determines whether small negative numbers in WQ variables should be zeroed.
• multi_ben: Flag to set WQ benthic fluxes to occur on all GLM layers, or just the bottom layer.
• wq_nml_file: Name of the input nml file to be read in for the WQ simulation.
27
Oct 2014
Running the model
The model may be run by navigating to the directory where the glm.nml file is and executing the model executable
glm.exe. The glm.exe file can be located in different directory and added to the system path if desired.
Windows users may wish to add the command into a glm.bat:
..\bin\glm.exe >glm.log
which will create a file that can simply be double-clicked from your file browser. The model will output to the NetCDF
and/or csv files, which can then be plotted in a number of ways.
Note that the Windows pre-compiled model executable is distributed in a 32-bit and 64-bit release; choose an
appropriate system.
Outputs and post-processing

The model includes several types of outputs, including the NetCDF file, an optional csv time-series file at a certain depth,
and an optional live contour plot as the model simulation runs, to enable the modeller to monitor simulation progress.
Live output plotting: plots.nml

If the model is run with the optional command line argument “--xdisp" then the model simulation will display live plots
of output parameters (Figure 8). The number of plots, parameters to plot and the colour bar limits are set in the file
plots.nml, which may be simply configured according to the input variables shown below:
&plots
nplots = 4
plot_width = 400
plot_height = 200
title = 'Temperature','Salinity','DO','extc'
vars = 'temp','salt','aed_oxygen_oxy',extc'
min_z = 0.0, 0.0, 0.0, 0.0
max_z = 30.0, 0.91, 400.0, 2.0
/
Table 6: Variables within the output.nc file available to be plot via plots.nml
Variable Name Description Units
temp Temperature °C
salt Salinity gL-1
rad Shortwave Radiation Wm-2
extc Extinction Coefficient m-1
dens Density kgm-3
uorb Orbital Velocity (@ Sediment-Water Interface) ms-1
taub Shear Stress (@ Sediment-Water Interface) Nm-2
<WQ> Any water quality model variable, e.g. aed_oxygen_oxy. various
Refer to keywords used for AED in Hipsey (2014), or output
summary information at the beginning of the simulation
28
Oct 2014
Sim ulation sum m ary: lake.csv
A daily summary of the simulation is summarised in the file lake.csv. This file includes lake scale information, related to
surface heating and cooling, the lake water balance and other relevant metrics. For outputs of specific simulated variables
at a particular depth, refer to the next sections.
Table 7: Summary information written to the lake.csv simulation output file.
Variable Column Description Units Note
date A Date yyyy-‐mm-‐dd
day B Julian day number -‐
Volume C Total lake volume ML (1000 m )

3

Tot Inflow Vol D Total daily inflow volume ML Sum of all inflow
Tot Outflow Vol E Total daily outflow volume ML Sum of all offtakes
Overflow Vol F Total daily volume of overflows ML Flows over the lake crest
Evaporation G Total daily volume of evaporation ML
Rain H Total daily volume of rainfall ML
Lake Level I Average lake level m

2
Surface Area J Lake surface area m
Blue Ice K Depth of blue ice m
Snow L Depth of snow m
White Ice M Depth of white ice m
Max Temp N Maximum daily temperature within lake °C
Min Temp O Minimum daily temperature within lake °C
Surface Temp P Surface temperature °C

-‐1
Daily Qsw Q Daily heat input from shortwave radiation MJ day
-‐1
Daily Qe R Daily latent (evaporative) heat lost from the lake MJ day
-‐1
Daily Qh S Daily sensible heat flux MJ day
-‐1
Daily Qlw T Daily net longwave flux MJ day
-‐2
Light U Incident light intensity µE m
Benthic Light V Percentage of the lake bottom exceeding 𝜙!"#!"#$ %
T W Average wave period s
Hs X Average significant wave height m
L Y Average wave length of surface wind waves m
LakeNumber Z Lake number -‐
Max dT/dz AA Maximum recorded vertical temperature gradient °C m-‐1
29
Oct 2014
Figure 8: Example of live output plots generated via the libplot library provided with the model.
Plotting in EXCEL
For simple time-series plots, the user can configure outputs from the model directly to a csv file for a certain depth
(defined relative to the bottom), and this information is defined in the glm.nml &output section. The columns to plot
must also be listed in this section and are user-definable. Users can choose to output at any frequency.
Plotting in M ATLAB
For more advanced or customised plots, then the user may load the output.nc NetCDF file into MATLAB. Recent
versions of MATLAB (MATLAB 2011a or after) natively support NetCDF and can load the file directly. An example MATLAB
script for plotting is shown below - this can be customised as required.
foldername = '../MyGLMSim/';
outname = '../ MyGLMSim /figures /';
mkdir([outname]);
data = nldncGLM([foldername,'/output.nc'])
varNames = names_netcdf([foldername,'output.nc']);
varsToPlot = varNames([20:64]);
for ii = 1:length(varsToPlot)
newFig = plotGLM(varsToPlot{ii},data);
figName = [outname,'/', varsToPlot {ii},'.png'];

print(gcf,'-dpng', figName,'-opengl');
close all
end
Plotting in R
The GLM output.nc NetCDF file can be read and plotted using the “R” package. A set of tools, “glmtools”, has been
developed in R by Jordan Read and Luke Winslow, and is available from: https://github.com/GLEON/glmtools. An
example plot from R is below (Figure 9).
30
Oct 2014
Figure 9: Image of temperature predicted by GLM, plotted using the R glmtools scripts.
Model Validation & Parameter Optimisation

There are numerous ways that model users may wish to assess model performance and adjust physical parameters in
glm.nml to optimise their calibration with observed data. As part of a GLEON (gleon.org) lake modelling working group, a
specific workflow for model assessment and parameter estimation has been trialled and is outlined below. The approach
a) uses the GLEON “LakeAnalyzer” analysis scripts to assess model comparisons across a range of metrics, and b)
combines these assessment scripts with a Markov Chain Monte Carlo (MCMC) method of parameter estimation.
Running the LakeAnalyzer validation

As part of a multi-lake comparison GLM has been compared against numerous different metrics of model performance.
These include simple measures like surface or bottom temperature, however it is also possible to compare the model’s
performance in capturing higher-order metrics relevant to physical limnology. To calculate these on the model and field
data, the LakeAnalyzer routines provided by Read et al. (2012). These have been adapted for GLM use and can be called
via the run using the calcGLMModelFit.m and plotGLMModelFit.m scripts. An example output from Lake Kinneret is
shown in Figure 10.
Field Files: model_fld_temp.wtr & model.bth
Together the model_fld_temp.wtr and the model.bth files give observed water temperature data and lake shape
details that are compared against the model output. Both files are comma separated text files in the same format as
required for running LakeAnalyzer.
The model_fld_temp.wtr file is a simple file consisting of time-stamped thermistor chain data. The first row given the
date and thermistor chain ID’s (Figure 11a). Note the date format must be saved as YYYY-MM-DD, prior to saving as a csv
file.
The model.bth file is a simple two column file consisting of each thermistors depth, and the area of the bathymetry at
that depth (Figure 11b). Save the files as a csv.
Running the M CM C param eter estim ation

As part of the modelling process, users may desire to adjust the GLM physical parameters to get the best fit with available
field data. GLM may be run with a Markov Chain Monte Carlo (MCMC) routine that can be used to provide improved
parameter estimates. On the GLM website we provide a version of GLM that works with the MCMC code provided by
Haario et al. (2006), though users may wish to develop their own optimisation approach.
The MCMC routines are available as MATLAB scripts that will call glm.exe during the run. This is run using the
runMCMC.m script. We have also prepared a pre-compiled form of the procedure that can run on Windows via the
command prompt, independent of MATLAB being installed, by running runMCMC.exe. Note that for the pre-compiled
31
Oct 2014
version that has been supplied, users must have the MATLAB runtime environment (MCR) installed
(http://www.mathworks.com.au/products/compiler/mcr/).
Observed data
GLM simulation
Figure 10: Example output from the GLM model assessment scripts.
Figure 11: Outline of the required field data files to run the model validation for
a) model_fld_tmp.wtr and b) model.bth.
32
Oct 2014
To run the model as part of the MCMC routine, users must prepare several extra files and directories. These include files
for providing the observed field data (as above) and also files associated with the MCMC routine:
• Field/model.bth
• Field/model_fld_temp.wtr
• InputFiles/glm_init.nml
• InputFiles/mcmc_config.nml
The model will run and output a log of model RMSE and other information about the performance of different parameter
combinations. These will be written to files in the Results/ folder.
MCMC files: glm_init.nml & mcmc_config.nml
The InputFiles/glm_init.nml file is simply a duplicate of the starting simulation GLM nml file. The structure of this
file is described in the above sections.
The InputFiles/mcmc_config.nml file is a namelist file which specifies certain parameters which governs how the
optimisation routine functions. Currently, there are four sections:
• &config
• &dataset
• &ssh
• &params
The &config section contains the following variables:
• Fld_temp_file – Path to model_fld_temp.wtr file

• Varname – GLM variable name of data within the wtr file (e.g. ‘temp’)
• Remote – Switch for whether routine is run locally (0) or an a server via ssh (1)
• Nsim_ini – Number of simulations to run to get base values
• Nsim_full – Number of optimisation simulations to run.
The &dataset section contains the following variables:
• Data_Subsets
• Model_Fit – Type of error calculation (e.g., RMSE, Nash-Sutcliffe, R2, etc)
The &ssh section contains the following variables:
• Host_Name
• Usr_Name
• Password
• Remote_Dir
• Run_GLM
• Output_File
• Varname
The &params section contains the following variables and these are those that are to be included in the parameter
optimisation. The values assigned to these variables is the starting parameter vector (see also Table 1):
• coef_mix_conv – Coefficient related to mixing efficiency of convective overturn.

• coef_wind_shear – Coefficient related to mixing efficiency of wind shear events.
• coef_mix_turb – Coefficient related to mixing efficiency of unsteady turbulence.
• coef_mix_kh – Coefficient related to Kelvin Helmholtz turbulent billows.
• coef_mix_hyp – Coefficient related to mixing efficiency of hypolimnetic turbulence
• ce – Bulk transfer coefficient for latent heat, CE.
• ch – Bulk transfer coefficient for sensible heat, CH.
• cd – Bulk transfer coefficient for momentum, CD.
33
Oct 2014
Examples & Support
Downloads & Further Support
To download the model, visit: http://aed.see.uwa.edu.au/research/models/GLM/
Support and FAQ’s are available at the Aquatic Ecosystem Modelling Network (AEMON) website:
http://sites.google.com/site/aquaticmodelling/
For specific development requests, please contact Dr Louise Bruce or A/Prof Matthew Hipsey
from the School of Earth and Environment, The University of Western Australia.
louise.bruce@uwa.edu.au
matt.hipsey@uwa.edu.au
Example Applications
Numerous applications are presented online as part of the GLM Multi-Lake Comparison Project (MLCP):
http://aed.see.uwa.edu.au/research/models/GLM/Pages/projects.html
Two example setups - “warmlake” and “coldlake” - are also available for download. These simulations demonstrate
working setups configured using various simulation options, including ice-cover for coldlake.
34
Oct 2014
References
Ashton, G.D. (ed.), 1986. River and lake ice engineering. Water Resources Publication.
Babanin, A.V. and Makin, V.K., 2008. Effects of wind trend and gustiness on the sea drag: Lake George study. Journal of Geophysical
Research: Oceans, 113(C2).
Bird, R.E., 1984. A simple, solar spectral model for direct-normal and diffuse horizontal irradiance. Solar energy, 32: 461-471.
Businger, J.A., Wyngaard, J.C., Izumi, Y., Bradley, E.F., 1971. Flux profile relationships in the atmospheric surface layer. Journal of
Atmospheric Sciences, 28: 181-189.
Condie, S.A. and Webster, I.T. 1997. The influence of wind-stress, temperature, and humidity gradients on evaporation from reservoirs.
Water Resources Research, 33: 2813-2822.
Dyer, A.J., 1974. A review of flux-profile relationships. Boundary-Layer Meteorology, 7: 363-372.
Fischer, H.B., List, E.G., Koh, R.C.Y., Imberger, J. and Brooks, N.H., 1979. Mixing in Inland and Coastal Waters. Academic Press.
Fischer et al., 1980. Chapter 6: Mixing in Reservoirs.
Francey, R.J., & Garratt, J.R., 1978. Eddy flux measurements over the ocean and related transfer coefficients. Boundary-Layer
Meteorology, 14: 153-166.
Gal, G., Imberger, J., Zohary, T., Antenucci, J., Anis, A., & Rosenberg, T., 2003. Simulating the thermal dynamics of Lake Kinneret.
Ecological Modelling, 162: 69-86.
Gill, A.E., 1982. Atmosphere-ocean dynamics (Vol. 30). Academic press.
Gu, R., and Stefan, H.G., 1993. Validation of cold climate lake temperature simulation. Cold regions science and technology, 22: 99-104.
Haario, H., Laine, M., Mira, A. and Saksman, E., 2006. DRAM: Efficient adaptive MCMC, Statistics and Computing 16: 339-354.
Hamilton, D.P. & Schladow, S.G. 1997. Water quality in lakes and reservoirs. Part I Model description. Ecological Modelling, 96: 91–110.
Harvey, L.D.D., 1990. Testing alternative parameterizations of lateral melting and upward basal heat flux in a thermodynamic sea ice
model, Journal of Geophysical Research, 95: 7359-7365.
Hicks, B.B. 1975. A procedure for the formulation of bulk transfer coefficients over water. Boundary Layer Meteorology, 8: 315-324.
Hicks, B.B., 1972. Some evaluations of drag and bulk transfer coefficients over water. Boundary Layer Meteorology, 3: 201:213.
Hipsey, M.R. and Sivapalan, M. 2003. Parameterizing the effect of a wind-shelter on evaporation from small waterbodies. Water Resources
Research, 39: 1339.
Hipsey, M.R., (ed.), 2014. Modelling aquatic dynamics with the AED modelling library. AED Report #30, The University of Western
Australia, Perth, Australia.
Hocking, G.C., and Patterson, J.C., 1991. Quasi-two-dimensional reservoir simulation model. Journal of Environmental Engineering, 117:
595-613.
Idso, S.B., and Jackson, R.D., 1969. Thermal radiation from the atmosphere. Journal of Geophysical Research, 74: 5397-5403.
Imberger , J., Patterson, J., Hebbert, B. and Loh, I.,1978. Dynamics of reservoir of medium size. Journal of the Hydraulics Division - ASCE,
104 No HY5: 725-743.
Imberger, J. and Patterson, J.C., 1981. A dynamic reservoir simulation model-DYRESM:5. In: H.B. Fisher (ed.), Transport Models for Inland
and Coastal Waters. Academic Press, New York: 310-361.
Imberger, J., and Patterson, J.C., 1990. Physical Limnology, p. 303-475. In: T. Wu (ed.), Advances in applied mechanics 27. Academic.
Imboden, D.M. and Wüest, A., 1995. Mixing Mechanisms in Lakes, p. 83-138. In: A. Lerman, D.M. Imboden and J.R. Gat (eds.), Physics and
Chemistry of Lakes. Springer-Verlag.
Jellison, R. and Melack, J.M. 1993. Meromixis and vertical diffusivities in hypersaline Mono Lake, California. Limnology and
Oceanography, 38: 1008–1019.
Ji, Z.G., 2008. Hydrodynamics and water quality: modeling rivers, lakes, and estuaries. John Wiley & Sons.
Kim, J-W., 1976. A generalized bulk model of the oceanic mixed layer. Journal of Physical Oceanography, 6: 686-695.
Kirk, J.T.O., 1994. Light and photosynthesis in aquatic ecosystems. Cambridge University Press.
Kleinhans, M.G. and Grasmeijer, B.T., 2006. Bed load transport on the shoreface by currents and waves. Coastal Engineering, 53 : 983-996.
Kraus, E.B. and Turner, J.S., 1967. A one-dimensional model of the seasonal thermocline: II The general theory and its consequences.
Tellus, 19: 98-106.
Laenen A. and LeTourneau A.P. 1996. Upper Klamath Lake nutrient loading study – Estimate of wind-induced resuspension of bed
sediment during periods of low lake elevation. US Geological Survey Open-File Report 95-414, 11 pp
Launiainen, J., 1995. Derivation of the relationship between the Obukhov stability parameter and the bulk Richardson number for flux-
profile studies. Boundary Layer Meteorology, 76: 165-179.
35
Oct 2014
Launiainen, J., and Cheng, B. 1998. Modelling of ice thermodynamics in natural water bodies. Cold Region Science and Technology, 27:
153-178.
Launiainen, J., and Vihma, T., 1990. Derivation of turbulent surface fluxes—An iterative flux-profile method allowing arbitrary observing
heights. Environmental Software, 5: 113-124.
Le Roux, J.P., 2007. A simple method to determine breaker height and depth for different deepwater wave height/length ratios and sea
floor slopes. Coastal Engineering, 54 : 271-277.
Luo, L., Hamilton, D., and Han, B., 2010. Estimation of total cloud cover from solar radiation observations at Lake Rotorua, New Zealand.
Solar Energy, 84: 501-506.
McKay, G.A., 1968. Problems of measuring and evaluating snow cover. In: Proceedings Workshop Seminar of Snow Hydrology.(Secretariat
Canadian National Committee for the IHD, Ottawa: 49-62.
McCord, S.A., and Schladow, S.G., 1998. Numerical simulations of degassing scenarios for CO2‐rich Lake Nyos, Cameroon. Journal of
Geophysical Research: Solid Earth, 103(B6): 12355-12364.
Mooij, W.M., Trolle, D., Jeppesen, E., Arhonditsis, G., Belolipetsky, P.V., Chitamwebwa, D.B.R., Degermendzhy, A.G., DeAngelis, D.L., De
Senerpont Domis, L.N., Downing, A.S., Elliott, A.E., Fragoso Jr, C.R., Gaedke, U., Genova, S.N., Gulati, R.D., Håkanson, L.,
Hamilton, D.P., Hipsey, M.R., Hoen, J., Hülsmann, S., Los, F.J., Makler-Pick, V., Petzoldt, T., Prokopkin, I.G., Rinke, K., Schep, S.A.,
Tominaga, K., Van Dam, A.A., Van Nes, E.H., Wells, S.A., Janse, J.H., 2010. Challenges and opportunities for integrating lake
ecosystem modelling approaches. Aquatic Ecology, 44: 633–667.
Monin, A.S., and Obukhov, A.M., 1954. Basic laws of turbulent mixing in the atmosphere near the ground. Jr. Akad. Nauk SSSR Geofiz.
Inst., 24: 163-187.
Patterson, J.C., Hamblin, P.F., and Imberger, J. 1984. Classification and dynamics simulation of the vertical density structure of lakes.
Limnology and Oceanography, 29: 845-861.
Patterson, J.C., and Hamblin, P.F., 1988. Thermal simulation of a lake with winter ice cover. Limnology and Oceanography, 33: 323-338.
Paulson, C. A., 1970. The mathematical representation of wind speed and temperature profiles in the unstable atmospheric surface layer.
Journal of Applied Meteorology, 9: 857-861.
Read, J.S., Hamilton, D.P., Jones, I.D., Muraoka, K., Winslow, L.A., Kroiss, R., Wu, C.H. and Gaiser, E., 2011. Derivation of lake mixing and
stratification indices from high-resolution lake buoy data. Environmental Modelling and Software, 26: 1325–1336.
Read, J., Hansen, G., Van Den Hoek, J., Hanson, P., Bruce, L.C., Markfort, C.D., 2014. Simulating 2368 temperate lakes reveals weak
coherence in stratification phenology. Ecological Modelling, 291: 142-150.
Rogers, C.K., Lawrence, G.A., and Hamblin, P.F., 1995. Observations and numerical simulation of a shallow ice-covered mid-latitude lake.
Limnology and Oceanography, 40: 374-385.
Sherman, F.S., Imberger, J. and Corcos, G.M., 1978. Turbulence and mixing in stably stratified waters. Ann. Rev. Fluid Mech. 10: 267-288.
Spigel, B., 1978. Wind mixing in Lakes. PhD thesis., University of California, Berkeley.
Spigel, R.H., Imberger, J. Rayner, K.N., 1986. Modeling the diurnal mixed layer. Limnology and Oceanography, 31: 533-556.
Strub, P.T. and Powell, T.M., 1987. Surface temperature and transport in Lake Tahoe: inferences from satellite (AVHRR) imagery.
Continental Shelf Research, 7: 1001-1013.
Swinbank, W.C. 1963. Longwave radiation from clear skies. Quarterly Journal of the Royal Meteorological Society, 89: 339-348.
Tennessee Valley Authority, 1972. Heat and mass transfer between a water surface and the atmosphere. Water Resources Research
Laboratory Report 14, Report No. 0-6803.
Tabata, S., 1973. A simple but accurate formula for the saturation vapour pressure over liquid water. Journal of Applied Meteorology, 12:
1410-1411.
Trolle, D., Hamilton, D.P., Hipsey, M.R., Bolding, K., Bruggeman, J., Mooij, W. M., Janse, J. H., Nielsen, A., Jeppesen, E., Elliott, J.E.,
Makler-Pick, V., Petzoldt, T., Rinke, K., Flindt, M. R., Arhonditsis, G.B., Gal, G., Bjerring, R., Tominaga, K., Hoen, J., Downing, A.S.,
Marques, D. M., Fragoso Jr, C.R., Søndergaard, M. and Hanson, P.C., 2012. A community-based framework for aquatic ecosystem
models. Hydrobiologia, 683: 25-34.
UNESCO. 1981. Technical papers in Marine Science. No. 36.
Vavrus, S.J., Wynne, R.H., and Foley, J.A., 1996. Measuring the sensitivity of southern Wisconsin lake ice to climate variations and lake
depth using a numerical model. Limnology and Oceanography, 41: 822-831.
Vickers, D., Mahrt, L., and Andreas, E.L., 2013. Estimates of the 10-m neutral sea surface drag coefficient from aircraft eddy-covariance
measurements. Journal of Physical Oceanography, 43: 301-310.
Weinstock, J., 1981. Vertical turbulence diffusivity for weak or strong stable stratification. Journal of Geophysical Research, 86(C10): 9925-
9928.
Wu J., 1973. Wind induced entrainment across a stable density interface. Journal of Fluid Mechanics, 61: 275-278.
Xenopoulos, M.A. and Schindler, D.W., 2001. The environmental control of near-surface thermoclines in boreal lakes. Ecosystems, 4:
699-707.
Yajima, H. and Yamamoto, S., 2014. Improvements of radiation estimations for a simulation of water temperature in a reservoir. Annual
Journal of Hydraulic Engineering, JSCE, 59, submitted.
36
Oct 2014
Yao, H., Samal, N.R., Joehnk, K.D., Fang, X., Bruce, L.C., Pierson, D.C., Rusak, J.A. and James, A., 2014. Comparing ice and temperature
simulations by four dynamic lake models in Harp Lake: past performance and future predictions. Hydrological Processes, 28:
4587–4601.
Yeates, P.S. and Imberger, J. 2004. Pseudo two-dimensional simulations of internal and boundary fluxes In stratified lakes and reservoirs.
International Journal of River Basin Research, 1: 1-23.
37
Oct 2014
Appendices
A: Bird solar radiation model

The Bird Clear Sky Model (BCSM) was developed by (Bird, 1984). This model predicts clear-sky direct beam, hemispherical
diffuse, and total hemispherical broadband solar radiation on a horizontal surface. Average solar radiation is computed
hourly with 10 user-specified input parameters (Table A1). The default parameters used in GLM are as outlined in this
table, but may be customised using the &bird parameter block in glm.nml.
Table A1: Parameters required for the BCSM model.
Value used to fit to data Value used in Luo et al.

Variable Description
from Perth, WA (2010) for Hamilton, NZ
Lat Latitude (+ for N) -31.77
Long Longitude (+ for E) 116.03
TZ Time Zone indicated by number of hours from GMT +7.5
AP Atmospheric Pressure (millibars) 1013
Oz Ozone Conc. (atm-cm) 0.279 0.279 - 0.324
W Total Precipitable Water Vapour (atm-cm) 1.5 1.1 - 2.2
𝐴𝑂𝐷!"" Aerosol Optical Depth at 500 nm 0.1 0.033 - 0.017
𝐴𝑂𝐷!"# Aerosol Optical Depth at 380 nm 0.15 0.038 - 0.019
𝛼!" Surface albedo 0.2 0.2
The solar constant in the model is taken as 1367 W/m2. This is corrected due to the elliptical nature of the earth’s orbit and
consequent change in distance to the sun. This calculation gives us the Extra-Terrestrial Radiation (𝜙!"# ), at the top of the
atmosphere:
𝜙!"# = 1367 1.00011 + 0.034221 cos Φ!"# + 0.00128 sin Φ!"# + 0.000719 cos Φ!"# (A1)
where the day angle, Φ!"# , is computed using, d, the day number:
𝑑−1
Φ!"# = 2𝜋 (A2)
365
The solar declination, Φ!"# (radians), is computed from:
0.006918 − 0.399912 𝑐𝑜𝑠 Φ!"# + 0.070257 𝑠𝑖𝑛 Φ!"# − 0.006758 𝑐𝑜𝑠 2 Φ!"# +
Φ!"# = (A3)
0.000907 𝑠𝑖𝑛 2Φ!"# − 0.002697 𝑐𝑜𝑠 3 Φ!"# + 0.00148 𝑠𝑖𝑛 3 Φ!"#
We then solve the equation of time:
0.0000075 + 0.001868 𝑐𝑜𝑠 Φ!"# − 0.032077 𝑠𝑖𝑛 Φ!"#

𝐸𝑄𝑇 = ×229.18 (A4)
−0.014615 𝑐𝑜𝑠 2 Φ!"# − 0.040849 𝑠𝑖𝑛 2 Φ!"#
in order to compute the hour angle, Φ!! , calculated with noon zero and morning positive as:
𝐸𝑄𝑇
Φ!! = 15 ℎ𝑟 − 12.5 + 𝐿𝑜𝑛𝑔 − 15 𝑇𝑍 + (A5)
4
where TZ is the time-zone shift from GMT. The zenith angle, Φ!"# (radians), is calculated from:
cos Φ!"# = cos Φ!"# cos Φ!! cos 𝐿𝑎𝑡 + sin Φ!"# sin 𝐿𝑎𝑡 (A6)
38
Oct 2014
When Φ!"# is less than 90°, the air mass factor is calculated as:
−1
0.15 (A7)
AM = 𝑐𝑜𝑠 Φ!"# + !.!"
93.885 − Φ!"#
which is corrected for atmospheric pressure, p (hPa),

𝐴𝑀 𝑝
𝐴𝑀! = (A8)
1013
AMP is then used to calculate the Rayleigh Scattering as:

!!.!"!# !"! !.!" ! !!!"! !!"! !.!" (A9)
𝑇!"#$%&'! = 𝑒
The effect of ozone scattering is calculated by computing ozone mass, which for positive air mass is:
!!.!"!# 0.002715 𝑂𝑧 𝐴𝑀 (A10)
𝑇!"!#$ = 1 − 0.1611 𝑂𝑧 𝐴𝑀 1 + 139.48 𝑂𝑧 𝐴𝑀 − !
1 + 0.044 𝑂𝑧 𝐴𝑀 + 0.0003 𝑂𝑧 𝐴𝑀
The scattering due to mixed gases for positive air mass is calculated as:
−0.0127 𝐴𝑀𝑝0.26 (A11)
𝑇!"# = 𝑒
Then the water scattering is calculated by getting the water mass:

𝑊𝑚 = 𝑊𝐴𝑀! (A12)
where W is the precipitable water vapour. This can be approximated from dew point temperature, eg.:
ln 𝑊 = 𝑎 𝑇! + 𝑏 (A13)
where a and b are regression coefficients which have been taken as 0.09, 0.07, 0.07 and 0.08 for values of a while b is 1.88,
2.11, 2.12 and 2.01 in spring, summer, autumn and winter (Luo et al., 2010).
Then the water scattering effect is calculated as:

2.4959 𝑊𝑚 (A14)
𝑇!"#$% = 1 −
1 + 79.034 𝑊𝑚 !.!"#" + 6.385 𝑊𝑚
The scattering due to aerosols requires the Aerosol Optical Depth at 380 nm and 500 nm:
𝑇𝑎𝑢𝐴 = 0.2758 𝐴𝑂𝐷!"# + 0.35 𝐴𝑂𝐷!"" (A15)
and the scattering due to aerosols is then calculated as:

−𝑇𝑎𝑢𝐴 0.873 1+𝑇𝑎𝑢𝐴−𝑇𝑎𝑢𝐴0.7088 𝐴𝑀0.9108
𝑇!"#$%$& = 𝑒 (A16)
We also define:
𝑇!! = 1 − 0.1 1 − 𝐴𝑀 + 𝐴𝑀!.!" 1 − 𝑇!"#$%$& (A17)
and:
0.5 1 − 𝑇!"#$%&'! + 0.84 1 − 𝑇!" (A18)
1 − 𝐴𝑀 + 𝐴𝑀!.!"
where the 0.84 value used is actually the proportion of scattered radiation reflected in the same direction as incoming
radiation.
The direct beam radiation on a horizontal surface at ground level on a clear day is given by,
𝜙!" = 0.9662 𝜙!"# 𝑇!"#$%&'! 𝑇!"!#$ 𝑇!"# 𝑇!"#$"% 𝑇!"#$%$& 𝑐𝑜𝑠 Φ!"# (A19)
𝜙!" = 0.79 𝜙!"# 𝑇!"!#$ 𝑇!"# 𝑇!"#$"% 𝑇!! 𝑐𝑜𝑠 Φ!"# (A20)
The total irradiance hitting the surface is therefore (W m-2):

𝜙!" + 𝜙!" (A21)
𝜙!" =
1 − 𝛼!" 𝛼!"#
The albedo is computed for the sky as:

𝑇!"#$%$&
𝛼!"# = 0.068 + 1 − 0.84 1− (A22)
𝑇!!
39
Oct 2014
B: Non-neutral bulk transfer coefficients
The iterative procedure used in this analysis is conceptually similar to the methodology discussed in detail in Launiainen
and Vihma (1990). The first estimate for the neutral drag coefficient is specified as a function of windspeed as it is has been
commonly observed that CDN increases with U10 (Figure A1). This is modelled by first by estimating:
0.001 𝑈!" ≤ 5
𝐶!"!!" = Option 1 : Francey and Garratt (1978), Hicks (1972) (A23)
0.001 1 + 0.07 𝑈!" − 5 𝑈!" > 5
𝐶!"!!" = 1.92×10!! 𝑈!"

!
+ 0.00096 Option 2 : Babanin and Makin (2008)
and then computing the Charnock formula with the smooth flow transition (e.g., Vickers et al., 2013):
𝛼𝑢∗! 𝜈 (A24)
𝑧! = + 0.11
𝑔 𝑢∗
!
where 𝛼 is the Charnock constant (0.012), u* is the friction velocity ( 𝐶!"!!" 𝑈!" ) using Eq A23, and the final drag is re-
computed using:
!
𝑘 (A25)
𝐶!"!!" = !"
ln
!!
where k is the von Karman constant. Note the neutral humidity/temperature coefficient, CHWN-‐10, is held constant at the
user defined CH value and doesn't scale with windspeed.
0.003
0.0025
0.002
CDN10
0.0015
0.001
0.0005
0
0 5 10 15 20 25
U10 (m/s)
Figure A1: Scaling of the 10m neutral drag coefficient with wind speed (Eqs. A23-25)
Under non-neutral conditions in the atmospheric boundary layer, the transfer coefficients vary due to stratification seen in
the air column, as was parameterised by Monin and Obukhov (1954) using the now well-known stability parameter, z/L,
where L is the Obukhov length defined as:
−𝜌! 𝑢∗! 𝜃! (A26)

𝐿=
! !"
𝑘𝑔 + 0.61
!! !
where 𝜃! = 𝜃 1 + 0.61𝑞 is the virtual temperature and H and E are the bulk fluxes. Paulson (1970) presented a solution
for the vertical profiles of wind speed, temperature and moisture in the developing boundary layer as a function of the
Monin-Obukhov stability parameter; the so-called flux-profile relationships:
𝑢∗ 𝑧 𝑧
𝑈! = ln − 𝜓! (A27a)
𝑘 𝑧! 𝐿
40
Oct 2014
𝜃∗ 𝑧 𝑧 (A27b)
𝜃! − 𝜃! = ln − 𝜓!
𝑘 𝑧! 𝐿
𝑞∗ 𝑧 𝑧 (A27c)
𝑞! − 𝑞! = ln − 𝜓!
𝑘 𝑧! 𝐿
where ψM, ψH and ψE are the similarity functions for momentum, heat and moisture respectively, and zo, zθ and zq are their
respective roughness lengths. For unstable conditions (L<0), the stability functions are defined as (Paulson 1970; Businger
et al., 1971; Dyer, 1974):
1+𝑥 1 + 𝑥! 𝜋 (A28a)
𝜓! = 2 ln + ln − 2 tan!! 𝑥 +
2 2 2
1 + 𝑥! (A28b)
𝜓! = 𝜓! = 2 ln
2
where
𝑧 ! !
(A29)
𝑥 = 1 − 16
𝐿
During stable stratification (L>0) they take the form:
𝑧 𝑧
−5 0 < < 0.5 (A30)
𝐿 𝐿
𝑧 !! 𝑧 !! 𝑧 𝑧
𝜓! = 𝜓! = 𝜓! = 0.5 − 4.25 −7 − 0.852 0.5 < < 10
𝐿 𝐿 𝐿 𝐿
𝑧 𝑧 𝑧
ln − 0.76 − 12.093 > 10
𝐿 𝐿 𝐿
Substituting equations (17)-(18) into (A27) and ignoring the similarity functions leaves us with neutral transfer coefficients as
a function of the roughness lengths:
𝑧 !! 𝑧 !!
𝐶!" = 𝑘 ! ln ln (A31)
𝑧! 𝑧!
where N denotes the neutral value and X signifies either D, H or E for the transfer coefficient and o, θ or q for the roughness
length scale. Inclusion of the stability functions into the substitution and some manipulation (Imberger and Patterson,
1990; Launianen and Vihma, 1990) yields the transfer coefficients relative to these neutral values:
𝐶! 𝐶!" 𝑘𝜓! 𝑘𝜓! 𝐶!" (A32)

= 1 + ! 𝜓! 𝜓! − −
𝐶!" 𝑘 𝐶!" 𝐶!"
Hicks (1975) and Launianen and Vihma (1990) suggested an iterative procedure to solve for the stability corrected transfer
coefficient using (A32) based on some initial estimate of the neutral value. The surface flux is subsequently estimated
according to (17-18) and used to provide an initial estimate for L (equation A26). The partially corrected transfer coefficient
is then recalculated and so the cycle goes. Strub and Powell (1987) and Launiainen (1995), presented an alternative based
on estimation of the bulk Richardson number, RiB, defined as:
𝑔𝑧 Δθ + 0.61 𝜃! Δq
𝑅𝑖! = (A33)
𝜃! 𝑈!!
and related as a function of the stability parameter, z/L, according to:
𝑧 k 𝐶!" 𝐶!"# − 𝜓!" (A34)

𝑅𝑖! = !
𝐿 𝑘 𝐶!" − 𝜓!
where it is specified that CHN = CWN = CHWN. Figure A2 illustrates the relationship between the degree of atmospheric
stratification (as described by both the bulk Richardson number and the Monin-Obukhov stability parameter) and the
transfer coefficients scaled by their neutral value.
41
Oct 2014
Figure A2: Relationship between atmospheric stability (bottom axis – z/L, top axis – Ri B ) and the bulk-transfer
coefficients relative to their neutral value (C X /C X N where X represents D, H or W) for several roughness values
(computed from Eq. A32). The solid line indicates the momentum coefficient variation (C D /C D N ) and the broken
line indicates humidity and temperature coefficient (C H W /C H W N ) variation.
42
Oct 2014

GLM v2 User Manual PDF

Uploaded by

Copyright:

Available Formats

GLM v2 User Manual PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

GLM v2 User Manual PDF

Uploaded by

Copyright:

Available Formats

What is the General Lake Model (GLM) designed for?

What is the General Lake Model (GLM) designed for?

What types of sites can be simulated with GLM?

What types of sites can be simulated with GLM?

2014

- General Lake Model

Model Overview and User Information

M.R. Hipsey, L.C. Bruce & D.P. Hamilton

Citing this manual:

Copyright © 2014. The University of Western Australia.

MODEL SCIENCE OVERVIEW ..................................................................................................................................... 7

SETUP & OPERATION ............................................................................................................................................... 23

EXAMPLES & SUPPORT ........................................................................................................................................... 34

REFERENCES ............................................................................................................................................................... 35

APPENDICES ............................................................................................................................................................... 38

Model Suitability & Data Requirements

𝑉! = 𝑉!!! + 𝐴!!! + 0.5(𝐴! − 𝐴!!! ) (ℎ! − ℎ!!! ) (1)

𝑄! = 𝑓!" 𝑓! 𝑅 − 𝑅! 𝐴!"# − 𝐴! (5)

Surface energy budget

Solar heating and light penetration

𝑓 𝐶 = 0.66182 𝐶 ! − 1.5236 𝐶 + 0.98475 (8)

Option 1 : Daily approximation, Hamilton & Schladow (1997)

Option 2 : Briegleb et al., (1986)

Option 3 : Yajima & Yamamoto (2014)

!.!! !!.!" !.!"# (9c)

where z is the depth of the layer from the surface, 𝑓!" is a 0

𝐴!"# = 𝐴! − 𝐴 ℎ!"# (11)

𝜙!"!"# = 𝜙!"!" − 𝜙!"!"# (13)

𝜙!"!" = 1 − 𝛼!" 𝜀!∗ 𝜎 𝑇! + 273.15 !

For latent heat:

Correction for non-neutral atmospheric stability

𝜙!! = 𝜌! 𝜆 𝛼! 𝐶! − 𝐶! (22)

Snow and ice dynamics

𝜙! 𝑇! + 𝜙!"# 𝑇! = 0 𝑇! < 𝑇!

𝜙!"# 𝑇! = 𝜙!"#$ − 𝜙!"#$% 𝑇! + 𝜙! 𝑇! + 𝜙! 𝑇! + 𝜙! 𝑇! (26)

Stratification and vertical mixing

Surface Mixed Layer:

𝐶! 𝑤∗! + 𝐶! 𝑢∗! Δt ≥ 𝑔!! z!"# + 𝐶! 𝑤∗! + 𝐶! 𝑢∗! ! !

𝑢!! 𝑧!"# + ∆𝜉 𝑢! 𝜉∆𝑢! 𝜉Δ𝑧!!! ∆𝜉 (37)

where the K-H length scale is 𝜉 = 𝐶!" 𝑢!! 𝑔!"

𝑔∆𝜌 𝑔 𝜌!!! − 𝜌!!! (40)

𝜀!"# ≈ 𝜀!"# = 𝐸!"#$ + 𝐸!"#$%& (41)

which is expanded and calculated per unit volume as:

0 ℎ! ≥ ℎ!"# − 𝑧!"# (43)

𝐶!! !/! (46)

𝐶!! 1 + 0.21 𝐶!! sin 𝛼!"# (47)

ℎ! = 1.2𝐸𝑑𝑥 + ℎ!!! (48)

1. That the lake shape approximates as an ellipse.

and wave height from:

The total shear stress at the lake bed is calculated as:

Figure 6: Slope, 𝝓𝒊𝒏𝒇 and half angle, 𝜶𝒊𝒏𝒇 of inflowing rivers.

Symbol glm.nml ID Description Units Default Reference Comments

𝜆 - Latent heat of evaporation J kg-1 2.453x106 Standard Not adjustable in glm.nml

𝜎 - Stefan-Boltzmann constant Wm K-2 -4

Mixing efficiency - shear Sherman et al. Best fit of experiments

Mixing efficiency - Kelvin- Sherman et al.

Mixing efficiency of hypolimnetic Weinstock General diffusivities in

Inflows & Outflows

𝐺 seepage_rate Seepage rate m day-1 0 Site specific

date A Date yyyy-‐mm-‐dd

day B Julian day number -‐

LakeNumber Z Lake number -‐

Max dT/dz AA Maximum recorded vertical temperature gradient °C m-‐1