Atmospheric & Oceanic Applications of Eulerian and Lagrangian Transport Modelling Joakim Kjellsson
Abstract
This thesis presents several ways to understand transports of air and water masses in the atmosphere and ocean, and the transports of energy they imply. It presents work using both various kinds of observations and computer simulations of the atmosphere and oceans. One of the main focuses is to identify similarities and differences between models and observations, as well as between different models. The first half of the thesis applies Lagrangian methods to study flows in the atmosphere and oceans. Part of the work focuses on understanding how particles follow the currents in the Baltic Sea and how they disperse. It is suggested that the commonly used regional ocean model for the Baltic Sea, RCO, underestimates the transport and the dispersion of the particles, which can have consequences for studies of e.g. biogeochemistry as well as for operational use. A similar methodology is used to study how particles are transported between the tropics and mid-latitudes by the large-scale atmospheric circulation. It is found that the mass transport associated with northbound and southbound particles can cancel in the zonally averaged circulation, and it is proposed that the degree of cancellation depends on the method of averaging. The latter half of the thesis focuses on Eulerian stream functions and specifically a thermodynamic stream function that combines the zonal and meridional circulations of the atmosphere into a single circulation. The stream function is used as a diagnostic to study the inter-annual variability of the intensity and thermodynamic properties of the global atmospheric circulation. A significant correlation to ENSO variability is found both in reanalysis and the EC-Earth coupled climate model. It is also shown that a set of models from the CMIP5 project show a slowdown of the atmospheric circulation as a result of global warming and associated changes in near-surface moisture content and upper-level radiative cooling.
Cover image: Plume of ash from the Eyjafjallajökull volcano seen by the Envisat Medium Resolution Imaging Spectrometer on 11 May 2010. Forecasts of ash clouds are often done by Lagrangian models. Photo: ESA c
Joakim Kjellsson, Stockholm 2014 ISBN 978-91-7447-823-5 Printed in Sweden by US-AB, Stockholm 2014 Distributor: Department of Meteorology, Stockholm University
Till studera
min
mor
meteorologi
Eva vid
dessutom
som
föreslog
Stockholms
stöttade
mig
att
jag
skulle
universitet,
hela
vägen
och . . .
Vi satte båtarna i bäcken såg dom flyta in i tunneln kanske vidare mot Vättern och kanalen ut till Nordsjön över vågorna mot Irland ut på havet och sen blåsa iväg och aldrig komma tillbaka mer till bäcken där det började
ur “Söndermarken” av Lars Winnerbäck
List of Papers
The following papers, referred to in the text by their Roman numerals, are included in this thesis. PAPER I: Surface drifters and model trajectories in the Baltic Sea, Kjellsson J. and Döös K. (2012), Bor. Env. Res., 17, 447–459.
PAPER II: Lagrangian decomposition of the Hadley and Ferrel Cells, Kjellsson J. and Döös K. (2012), Geophys. Res. Lett., 39, L15807. DOI: 10.1029/2012GL052420 PAPER III: The Atmospheric General Circulation in Thermodynamical Coordinates, Kjellsson J., Döös K., Laliberté F. and Zika J. (In press.), J. Atmos. Sci. DOI: 10.1175/JAS-D-13-0173.1 PAPER IV: Slowdown of Atmospheric General Circulation with Global Warming, Kjellsson J. (Manuscript),
Reprints were made with permission from the publishers Boreal Environment Research Publishing Board (Paper 1), John Wiley & Sons Ltd. (Paper 2), and the American Meteorological Society (Paper 3).
Author’s contribution
Paper 1 emerged within the EU/BONUS+ project BalticWay where my task was to validate the upper-ocean mass transport in an ocean model using observations collected by Kristofer Döös. The idea came from me and Kristofer Döös. Ocean model integrations were provided by Markus Meier from SMHI. I processed the observational data, designed and performed all trajectory simulations and all analyses. The paper was written by me and Kristofer Döös. For Paper 2, the idea came from discussions between me, Kristofer Döös, Jonas Nycander and Rodrigo Caballero. I designed the trajectory simulations together with Kristofer Döös. I performed all trajectory simulations, analyses and wrote the paper with inputs from Kristofer Döös. Paper 3 was an idea based on discussions with me, Kristofer Döös, Johan Nilsson, Maxime Ballarotta, Jan Zika, and Frédéric Laliberté partly during an oceanography workshop in Stockholm. EC-Earth integrations were provided by Laurent Brodeau. I performed all analyses and wrote the initial draft of the paper. The final paper was improved by me, Kristofer Döös, Jan Zika and Frédéric Laliberté. Paper 4 was my own idea based on results from Paper 3. I collected all data, performed all analyses, and wrote the paper with inputs from Kristofer Döös and Maxime Ballarotta.
Contents
Abstract
ii
List of Papers
v
Author’s contribution
vii
Acknowledgements
xi
1
What is Lagrangian and what is Eulerian?
13
2
The Eulerian framework 2.1 Computer models of atmosphere and ocean . . . . . . . . . . . . . . 2.2 Eulerian stream functions . . . . . . . . . . . . . . . . . . . . . . . .
17 17 19
3
The Lagrangian framework 3.1 Lagrangian observations of the world oceans . . . . . . . 3.2 TRACMASS - An algorithm for Lagrangian trajectories 3.3 Some Lagrangian statistics . . . . . . . . . . . . . . . . 3.4 Lagrangian stream functions . . . . . . . . . . . . . . .
23 23 23 26 27
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
4
Transports in the Baltic Sea and environmental risks
29
5
The overturning circulation of the atmosphere 5.1 Past and present . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31 31 35
6
Outlook
39
Sammanfattning References
xli xliii
Acknowledgements
They say that behind every successful meteorologist stands an even more successful oceanographer. In my case, there are two: Kristofer Döös and Jonas Nycander, who decided to hire me when others would not. Thank you for giving me the opportunity to pursue a PhD in Atmosphere and Ocean sciences! I’d also like to thank the people from the Baltic Way project for support, especially the leaders Tarmo Soomere and Ewald Quak. I’m also grateful for all the help and advice from Peter Lundberg and Johan Nilsson over the years. Thank you! Also, thank you Tom Rossby and my ship mates onboard the R/V Endeavour from University of Rhode Island for teaching me how observational oceanography is done. I’d like to thank Fred Laliberté and Jan Zika, two brilliant young scientists from whom I learned a lot during the latter half of my PhD, and also Bror Jonsson (without the ö. . . ) for introducing me to Princeton, being so supportive, and for all the long discussions of science, cold beverages and politics. I also wish to thank the Climate Research School and the Bolin Centre for Climate Research for the great courses, the travel grants and the PhD conference. I’ve also had a lot of help from the National Supercomputer Centre at Linköping University and their fantastic facilities and support team. Furthermore, I thank Rune Grand Graversen et al. for a great CESM course in Stockholm. Thanks to my travel companions over the years for making conferences such as EGU and AGU way more fun than they seem. Thank you Maxime (and Natalie), Magnus, Jenny, Cecilia, Marie, Cian, Abubakr, Peggy, Matthias and Henrik! And thank you Susanne for helping me filling out the travel bills! Thanks also to the MGFs Saeed, Laurent and Cian for all the lunches. A big thanks to the innebandy team (a member list to long for the purposes of this paragraph) for making Thursdays my favourite working day. A very special thanks to my roommate Léon. After 11 years together in the Swedish educational system we now part ways. Your hard work has been a true inspiration, and I wish you all the best of luck in your future career as a scientist, marathon runner and family man. Moreover my family are worthy of a special acknowledgement. I would not have survived the undergraduate studies without the support from my mom Eva and dad Tommy. Nor would it not have been this fun without my sister Sofia. The greatest thanks of all goes out to my dear girlfriend Viktoria. Thank you for putting up with
me when I try to explain my work, and thank you for supporting me when things did not go that well. PS. Climate-model data for Paper 4 was partly gathered during a time when most websites for US research centers were down (PCMDI, GFDL, NASA, etc.), so I’d like to “thank” President Obama, the US Congress and Senate for making data collection a challenge.
1. What is Lagrangian and what is Eulerian?
When you observe a fluid in motion, for instance water in a river, there are two possible viewpoints: the Eulerian1 and Lagrangian2 view. In the Eulerian view the observer stands at a fixed point in space and observes how the flow evolves. In the Lagrangian view the observer moves with the flow and observes its evolution while moving. A simple analogy is the various ways of seeing the “Vasaloppet”, the 90 km cross-country skiing race between Sälen and Mora held every year in Sweden. Bystanders will see the race from a fixed point in space, i.e. Eulerian, while TV cameras follow the race by moving with the leader, i.e. Lagrangian. Consider the temperature, T , at Observatoriekullen in Stockholm. The observer is then taking the Eulerian view (fixed point in space), and the temperature in Stockholm depends both on local effects (i.e. the sun warming the air) and advective effects (i.e. warm air blowing in over Stockholm). However, if the observer would take the Lagrangian view (follow the air as it moves in over Stockholm) the total effect on the air mass would be observed. We summarise this into the equation ∂T + v · ∇T = |∂t {z } Eulerian
dT , dt |{z}
(1.1)
Lagrangian
where v = (u, v, w) is the three-dimensional wind (cf. Vallis (2006)). The Lagrangian derivative is thus a sum of the local Eulerian effect (∂ T /∂t) and the advective Eulerian effect (v · ∇T ). The word “Lagrangian” is sometimes replaced by “material” to indicate that the observer is following a piece of material, i.e. a particle or a parcel. Observations of the atmosphere and oceans are often from an Eulerian view, i.e. weather stations, satellites, moored bouys, etc. The weather forecasts presented on television is often described using the Eulerian view, i.e. “the temperature in Stockholm will be +5◦ C” and “the wind will be 5 m s−1 westerly”. Observations from a Lagrangian view are also used, but to a much smaller extent. An example is the Global Drifter Program3 which uses about 1000 satellite-tracked floating buoys in the ocean which can measure temperature and air pressure every hour while drifting with the ocean currents. Another example is Morel & Desbois (1974) who released about 1 After
Leonhard Euler (1707-1783) Joseph Louis Lagrange (1736-1813) 3 http://www.aoml.noaa.gov/phod/dac/index.php 2 After
13
Trajectory Forecast Mississippi Canyon 252
NOAA/NOS/OR&R Estimate for: 0600 CDT, Thursday, 5/06/10 Date Prepared: 1300 CDT, Wednesday, 5/05/10
This forecast is based on the NWS spot forecast from Wednesday, May 5 AM. Currents were obtained from the NOAA Gulf of Mexico, Texas A&M/TGLO, and NAVO/NRL models and HFR measurements. The model was initialized from satellite imagery analysis provided by NOAA/NESDIS obtained Wednesday morning and a Wednesday morning helicopter overflight. The leading edge may contain tarballs that are not readily observable from the imagery (hence not included in the model initialization).
Mobile Gulfport Bay St Louis Mississippi Sound
Milton Pensacola
Pascagoula Mobile Bay
Forecast location for oil on 06-May-10 at 0600 CDT
Chandeleur Sound
Breton Sound
F G Mississippi Canyon 252 Incident Location
0
15
30
60
Miles
Winds are forecast to be light (5 kts) and variable continuing through Thursday morning. S/SE winds at 10 kts are expected to resume again Thursday afternoon/evening and continue through Friday. The Mississippi Delta, Breton Sound and Chandeleur Sound continue to be threatened by shoreline contacts throughout the forecast period. All three ocean models show a westward current developing south of the Delta transporting oil to the west of the Delta.
Trajectory Uncertainty Light Medium Heavy Potential X beached oil
Next Figure 1.1: Forecasted transport from the Deepwater Horizon oil leak calculated by the Forecast: National Oceanic and Atmospheric Administration (NOAA) using a Lagrangian transport May 6th AM model. Blue colours show the concentration of oil, and red marks show locations where oil may reach the shore. Image courtesy of NOAA’s Office of Response and Restoration.
14
500 balloons over the Southern Hemisphere and used information about their motion to map the winds around Antarctica. Lagrangian forecasts can also be of great importance. In 2010, the Deepwater Horizon oil ridge in the Gulf of Mexico sank causing the largest oil leak in U.S. history with devastating consequences for the marine life and coastal areas. The National Oceanic and Atmospheric Administration (NOAA) produced daily forecasts of the oil spreading to warn the public and to plan operations to recover or burn some of the oil still at sea.
15
16
2. The Eulerian framework
2.1
Computer models of atmosphere and ocean
Almost all computer models of the atmosphere and oceans are Eulerian, i.e. they give the results from a fixed point of view. To represent the global atmosphere in a computer one must first discretise the atmosphere by dividing it into a number of boxes as in Fig. 2.1. The method is very similar to dividing a photograph into pixels. A photograph on a computer is actually a digital representation of the view from the camera lens. In the same way, the global atmosphere in a computer model is a digital representation of a continuous atmosphere. Similar to photography, more pixels/boxes should generally give a picture/model that is more similar to reality. However, a higher resolution, i.e. more boxes, consumes more computer resources. As meteorologists and climate modellers strive to produce better weather forecasts and climate predictions some of the most powerful computers in the world are used for running atmospheric, ocean, or climate models. Fundamentally, a computer model of the atmosphere is a computer program that solves a series of equations for each box (Fig. 2.1). The output from an atmospheric model is thus temperature, wind, humidity, pressure, etc. in each box. An ocean model typically outputs current velocity, temperature, salinity, density, etc. A “climate” model is a coupled model, where models of e.g. atmosphere, ocean, sea-ice, land, air chemistry, marine ecosystem, etc., run at the same time and feed back on each other. It therefore outputs a wide range of data. A model of the atmosphere, like the Integrated Forecasting System (IFS) model from the European Centre for Medium-range Weather Forecasts (ECMWF), solves an equation similar to ∂T ∂T κTv ω + v · ∇T + η˙ − = PT + KT , ∂t ∂ η (1 + (δ − 1)q)p | {z } | {z } physics
(2.1)
dynamics
in each grid box. Similar equations for wind, humidity, etc. can be found in e.g. Kalnay (2003). In this equation, T is the temperature and the terms on the left-hand side correspond to: local heating/cooling, horizontal transport, vertical transport and adiabatic expansion. These terms are often called the “dynamics” of a model since they represent the evolution of the flow using basic fluid dynamics. The two terms on the right-hand side, PT and KT , are the “parameterisation” and “diffusion” terms, often called the “physics” of the model. These include the effect of processes that are not resolved in computer models. 17
Part IV: Physical Processes
Chapter 1
Overview Table of contents 1.1 1.2
Introduction Overview of the code
1.1
INTRODUCTION
a)
top of atm. z ≈ 40 km, p = 0 hPa z
b)
The physical processes associated with radiative transfer, turbulent mixing, subgrid-scale orographic drag, moist convection, clouds and surface/soil processes have a strong impact on the large scale flow of the atmosphere. However, these mechanisms are often active at scales smaller than the horizontal grid size. Parametrization schemes are then necessary in order to properly describe the impact of these subgridscale mechanisms on the large scale flow of the atmosphere. In other words the ensemble e↵ect of the subgrid-scale processes has to be formulated in terms of the resolved grid-scale variables. Furthermore, forecast weather parameters, such as two-metre temperature, precipitation and cloud cover, are computed by the physical parametrization part of the model.
This part (Part IV ‘Physical processes’) of the IFS documentation describes only the physical parametrization package. After all the explicit dynamical computations per time-step are performed, the physics parametrization package is called by the IFS. The physics computations are performed only in the vertical. The input information for the physics consists of the values of the mean prognostic variables (wind components, temperature, specific humidity, liquid/ice water content and cloud fraction), the provisional dynamical tendencies for the same variables and various surface fields, both fixed and variable.
c)
p surface z = 0 km, p ≈ 1000 hPa
Figure 1.1 Schematic diagram of the di↵erent physical processes represented in the IFS model.
Figure 2.1: The “grid” of a computer model of the atmosphere. a) The grid as seen from grid for a given square in a). Each box is3 called a “grid box” and represents a region enclosed between two longitudes, two latitudes and two vertical levels. c) Each grid box includes a set of processes that are too small to be resolved and are instead “parameterised”. Figure c) courtesy of ECMWF.
IFS Documentation – Cy31r1 above. b) The vertical
18
The letter “P” in PT denotes “Parameterisation” which means adding the effect of a process rather than the process itself. A few examples of “parameterised” processes in the IFS atmospheric model from ECMWF is shown in Fig. 2.1c. For instance, clouds are not captured by any global weather forecasting model, but since we know the effect they have we can add a parameterisation that explicitly changes the humidity, temperature, etc. in the model in accordance with our knowledge of clouds. The parameterisations, PT , also includes heating from short-wave radiation from the Sun, latent heat release, absorption by greenhouse gases, and more. The term KT represents mixing not captured by the model, also known as “diffusion”. In general, atmosphere and ocean models include more diffusion than what would be realistic in order to keep the flow smooth. Similar to the equation for temperature (eq. (2.1)) an atmospheric model also includes equations for e.g. wind, pressure and humidity, while an ocean model includes equations for e.g. currents and salinity. These equations also have some specific parameterisations as shown in Fig. 2.1c. For instance, atmosphere and ocean models always parameterise the effect of sub-grid turbulence to some extent. By sub-grid turbulence, we mean motions that are too small to be resolved by the model, i.e. smaller than the grid boxes. An example is the friction caused when air blows over the Earth’s surface. Sub-grid turbulence is especially important in the ocean since geostrophic eddies1 can not be resolved by most global ocean models and must be parameterised. The earliest ocean models did not include such a parameterisation which meant that they had severe difficulties in producing a realistic global climate. Gent & McWilliams (1990) were the first to include this which greatly improved the realism of their ocean model. This parameterisation is nowadays known simply as “GM” and is still frequently used.
2.2
Eulerian stream functions
There a many ways to analyse data given in the Eulerian framework, e.g. maps of temperature and winds in the atmosphere or currents in the ocean. A common method to analyse the Eulerian transports of mass and energy in the atmosphere and oceans is to use stream functions. Mathematically, a stream function is a two-dimensional function, Ψ(x, y), defined as ∂Ψ = v, ∂x
−
∂Ψ = u, ∂y
(2.2)
where u and v are the east–west and north–south velocities respectively. If the flow is two-dimensional and does not change in time, then the stream lines in the stream function Ψ are equal to Lagrangian trajectories. However, the atmosphere and oceans are three-dimensional flows that vary in time, and must therefore be reduced to two 1 similar
to the low- and high-pressure systems in the atmosphere
19
dimensions. A barotropic stream function, Ψ(x, y), can be obtained by integrating vertically and averaging over time, ∂Ψ 1 = ∂x t1 − t0
Z t1 Z z
−
v dz dt, t0
0
∂Ψ 1 = ∂y t1 − t0
Z t1 Z z
u dz dt.
(2.3)
0
t0
The unit of Ψ(x, y) is [m3 s−1 ] which means a transport of volume. The density of sea water is fairly constant in the ocean so a volume transport of 1 m3 s−1 is a mass transport of ≈ 1000 kg s−1 . For this reason, the unit Sverdrup 1 Sv = 106 m3 s−1 is often taken as a volume transport in the ocean but as a mass transport of 1 Sv = 109 kg s−1 in the atmosphere1 . The case for the atmosphere is discussed further below. The depth-integrated barotropic stream function, Ψ(x, y), represents the total transport projected onto the longitude–latitude plane. We can instead integrate over longitudes in which case we flow reduces to latitude–depth coordinates, 1 ∂Ψ = ∂z t1 − t0
Z t1 Z xE
v dx dt, t0
xW
−
1 ∂Ψ = ∂y t1 − t0
Z t1 Z xE
w dx dt, t0
(2.4)
xW
where xE and xW are the eastern and western integral bounds. The stream function Ψ(y, z), also known as the meridional overturning stream function, is a commonly used tool to study the north–south transports of mass and energy in the atmosphere and ocean. Calculating the meridional overturning stream function for the Atlantic ocean shows the Atlantic Meridional Overturning Circulation (AMOC) and how warm water masses are transported from the tropical Atlantic to the North Atlantic (cf. Kuhlbrodt et al. (2007)) which plays a large role in controlling European climate. In the atmosphere, the meridional overturning stream function is often shown using pressure instead of height as a vertical coordinate, i.e. Ψ(y, p). Using pressure coordinates and multiplying the velocity by 1/g where g = 9.81 m s−2 is the gravitational acceleration gives the stream function in units [kg s−1 ]. It is thus a mass transport rather than a volume transport as in the ocean. Several recent studies have generalised the meridional overturning stream function so that any variable can be used as a vertical coordinate but still yield the result as a mass transport (Döös & Nilsson, 2011; Pauluis et al., 2008). The meridional overturning stream function with a generalised vertical coordinate, Ψ(y, χ), is 1 ∂ Ψ(y, χ) = ∂χ t1 − t0
Z t1 I Z ps t0
x 0
δ [χ − χ 0 (x, y, p)]g−1 v d p dx dt,
(2.5)
where the Dirac function, δ [χ − χ 0 ] is a function where δ = 1 if χ 0 = χ and δ = 0 otherwise. Note that we integrate over all longitudes. Multiplying the velocity by the Dirac function and integrating is thus a way to “search” all longitudes and pressures at y and only “select” velocities where χ 0 (x, y, p) = χ. The solution for Ψ(y, χ) is 1 Density
20
of sea water is ρ ≈ 1000 kg m−3 so 1 m3 s−1 ≈ 1000 kg s−1
1 Ψ(y, χ) = t1 − t0
Z t1 I Z ps t0
x 0
µ[χ − χ 0 (x, y, p0 ,t)]g−1 v d p dx dt,
(2.6)
where µ is a Heaviside function which is the integral of the Dirac function and thus µ = 1 when χ 0 ≤ χ and µ = 0 otherwise. Examples of meridional overturning stream functions calculated using eq. (2.6) are shown in Fig. 5.2. A further generalisation of eq. (2.6) is to use two generalised coordinates. This has been done recently to study the atmospheric circulation from a purely thermodynamic perspective (Kjellsson et al., 2013; Laliberté et al., 2013). This generalisation is done by taking eq. (2.5) and using λ as a second general coordinate. ∂ Ψ(λ , χ) = ∂χ
Z Ω
δ [χ − χ 0 (x, y, p)]δ [λ − λ 0 (x, y, p)]g−1~v dΩ,
(2.7)
where integrating over Ω is equal to integrating over the full three-dimensional global atmosphere. Multiplying by two Dirac functions allows for a “selection” of the three-dimensional velocities where λ 0 = λ and χ 0 = χ. To calculate the stream function, Ψ(λ , χ) and taking temporal variations into account, we get Ψ(λ , χ) =
1 t1 − t0
Z Ω
δ [λ − λ 0 (x, y, p,t)]µ[χ − χ 0 (x, y, p,t)]g−1~v dΩ.
(2.8)
This method was used in Papers 3 and 4 of this thesis to define the hydrothermal stream function which uses latent heat and dry static energy as coordinates. The discretised form of eq. (2.8) is shown in Fig 11 of Paper 3 of this thesis.
21
22
3. The Lagrangian framework
3.1
Lagrangian observations of the world oceans
As briefly mentioned in the first chapter, the Global Drifter Program gathers data from some 1000 satellite-tracked buoys in the world oceans. These buoys float in the uppermost layer of the ocean and drift freely with the horizontal currents. Hence, the name surface drifters. The WOCE1 standard design of a surface drifter is a float at the surface containing a battery, GPS chip and in some cases a thermometer and/or a barometer (Sybrandy et al., 2009). Attached to the float is a 12 meter tether line that connects to a hollow drogue. The drogue is 6 meter long and the drogue thus sits between 12 and 18 m depth. This design ensures that the float at the surface actually follows the currents at 12 − 18 m depth to a high accuracy. An overview of the design is given by Niiler et al. (1995) and Lumpkin & Pazos (2007). Each surface drifter transmits its position and, if possible, sea-surface temperature (SST) and air pressure once every hour to a satellite. The positions can then be used to calculate the velocity of near-surface currents in the world oceans. Several studies have compared the surface-drifter data to ocean-model simulations to evaluate the realism of the models. Some studies have used the drifter data to obtain gridded Eulerian fields of velocity, eddy kinetic energy and other metrics (Garraffo et al., 2001; Rupolo, 2007). Others have used an algorithm to simulate Lagrangian trajectories using the ocean model velocity fields (Döös et al., 2011; Lumpkin et al., 2002; McClean et al., 2002) and compared the trajectories and their statistics. In Paper 1 of this thesis, surface drifter data from the Baltic Sea is used to evaluate the realism of the near-surface currents in the widely used Rossby Centre regional Ocean climate (RCO) model for the first time ever.
3.2
TRACMASS - An algorithm for Lagrangian trajectories
Paper 1 and Paper 2 of this thesis applies the TRACMASS Lagrangian trajectory code to solve atmospheric and oceanic problems. A description of the TRACMASS algorithm and other trajectory codes is therefore presented here. 1 World
Ocean Circulation Experiment
23
straightforward when using a C-grid model. Although B-grid velocities just need to be projected on the C-grid by making a meridional average of two zonal velocities (uC i,j,k = B 0.5(uB i,j,k + ui,j
viB 1,j,k ))
1,k ))
C B + and a zonal average of two meridional velocities (vi,j,k = 0.5(vi,j,k
for each grid box.
vi,j
yj
Latitude
(x1,y1)
ui-1,j
ui,j
(x0,y0)
yj-1 xi-1
vi,j-1
xi
Longitude
Figure 3.1: The trajectory of a TRACMASS particle passing through a grid box similar to those in Fig. of 2.1. grid box is seen above one where v j and aremodel the Figure 1.2:depicted Illustration a The trajectory [x(t), y(t)]from through grid box.v j−1 The meridional at the the northern velocities arevelocities defined at walls of and the southern grid box.walls, and ui and ui−1 are the velocities at the eastern and western walls. In a finite difference model there is no information of scales below the grid size. The tracers are regarded as homogeneous within each grid box and the velocities are only
To calculate the side pathwalls. of a particle, e.g. as possible done by toNOAA Fig. 1.1, we must defined on the grid It is, however, defineinthe velocity inside a
know the flow velocity and the origin of the particle. The path along which the particle moves is called a Lagrangian trajectory. The principle is to use the fact that velocity is the time-derivative of the position, i.e.
grid box by interpolating linearly between the discretised velocity values of the opposite
dx , dt which means that the position, x, of a particle at time t can be found by u=
x = x0 +
(3.1)
Z t
u dt.
(3.2)
t0
Numerically, the velocity at time step n, un , is used to calculate the displacement from xn to xn+1 , xn+1 = xn + un ∆t,
(3.3)
where ∆t is the time step, and n is the time index. This method is often called Euler’s method and is common for solving Ordinary Differential Equations. The method is however non-centered in time and is only first order accurate. The accuracy can be doubled by using a second order Runge-Kutta scheme. xn+1 u
n+1/2
= xn + un+1/2 ∆t,
(3.4)
= u (x + 0.5u ∆t,t + 0.5∆t) , n
n
n
(3.5)
where un+1/2 is the velocity between time step n and n + 1. A scheme similar to this is used by NOAA’s HYSPLIT model (Draxler & Hess, 1998). A popular scheme to use is the 4th order Runge-Kutta scheme where the accuracy is again doubled 24
u1
1 = xn + ∆t(u1 + 2u2 + 2u3 + u4 ), 6 = un ,
u2
= u(x + 0.5u1 ∆t,t + 0.5∆t),
(3.8)
u3
= u(xn + 0.5u2 ∆t,t n + 0.5∆t),
(3.9)
u4
= u(x + u3 ∆t,t + ∆t).
xn+1
n
n
n
n
(3.6) (3.7)
(3.10)
This scheme, sometimes denoted RK4, makes several approximations of the velocity at time step n, n + 1/2 and n + 1 to calculate the next position xn+1 . The TOMCAT/SLIMCAT Chemical Transport Model (Chipperfield, 2006) as well as the model used by Bowman & Carrie (2002) use an RK4 scheme to calculate the trajectories of particles. The TRACMASS1 code (Blanke & Raynaud, 1997; Döös, 1995; Döös et al., 2013) is based on equations that look quite different. Start with the mass flux in a grid box Ui = ui ∆y∆z.
(3.11)
The continuous mass flux U(x) is then found by linear interpolation U(r) = Ui−1 + (r − ri−1 )(Ui +Ui−1 ),
(3.12)
where the position x is exchanged for r = x/∆x. Eq. (3.12) can be set up as a differential equation dr + β r + δ = 0, ds
(3.13)
where β = Ui−1 − Ui and δ = −Ui−1 − β ri−1 and time t is exchanged for s = t/(∆x∆y∆z). Eq. (3.13) has the solution to r and s, δ δ e−β (s−s0 ) − r0 + β β 1 r1 + δ /β = s0 − log , β r0 + δ /β
r(s) = s1
(3.14) (3.15)
where r(s) is the position as a function of the time coordinate s, and s1 − s0 is the time it takes to travel from r0 to r1 . The TRACMASS algorithm thus gives the position and time for each trajectory by evaluating eqs. (3.14) and (3.15) which are exact for a stationary flow. de Vries & Döös (2001) developed an exact solution for TRACMASS trajectories in time-varying flows. This is opposed to Runge-Kutta methods where the position is approximated using e.g. eqs. (3.6)-(3.9). 1 TRACing
the water MASSes of the North Atlantic and the Mediterranean
25
3.3
Some Lagrangian statistics
Paper 1 focuses on calculating and comparing Lagrangian statistics for trajectories of surface drifters and model-simulated (“synthetic”) trajectories in the Baltic Sea to evaluate the realism of the model. There exists several kinds of statistical metrics to diagnose the trajectories of particles in a flow. An overview was given by LaCasce (2008). Two commonly used metrics are absolute and relative dispersion. Absolute dispersion, denoted D2A (t), is a measure of the displacement from the origin, D2A (t) =
1 M ∑ |x(t) − x(0)|2 , M m=1
(3.16)
where M is the number of particles. Absolute dispersion is often denoted as a squared distance with units [m2 ]. It can be calculated for a single particle and is thus sometimes called single-particle dispersion. Relative dispersion, denoted D2R (t), is a measure of the distance between two particles at a given time, D2R (t) =
1 |xi (t) − x j (t)|2 , N p i6∑ =j
(3.17)
where N p is the number of pairs of particles such that particle i is never the same as particle j. Relative dispersion is often calculated for pairs of particles that are initially very close to study how they separate. This gives information about the diffusivity of the flow, i.e. rate of mixing (Klocker et al., 2012; LaCasce & Bower, 2000; Sallée et al., 2008). The variability of the flow can be diagnosed by calculating the auto-correlation, R(τ), of the “eddy” velocity, u0 , defined as the total velocity minus the time-mean velocity, u0 (t) = u(x) − u. The velocity auto-correlation at time lag τ is the correlation of the eddy velocity time series u0 (t) and the eddy velocity time series u0 (t + τ) which has been shifted in time by τ. Hence, 1 T →∞ σ 2 T V
R(τ) = lim
Z T 0
u0 (t + τ) · u0 (t) dt,
(3.18)
where σV2 is the variance of the eddy velocity, u0 . The velocity auto-correlation, R(τ), can be said to hold the “memory” of the particle. The Lagrangian velocity time scale, TL , can be seen as a measure of how long time a particle “remembers” when it has been and is defined as TL =
Z ∞
R(τ) dτ.
(3.19)
0
If the flow is highly turbulent the velocity of a particle will de-correlate quickly and TL becomes short. If, on the other hand, the flow is less variable, the velocity will 26
be auto-correlated for a much longer time and TL will be long. When calculating TL it is common to filter out variations on short time scales e.g. inertial oscillations or tides since they tend to dominate R(τ) otherwise. It is often impractical to evaluate the full integral in eq. (3.19). Lumpkin et al. (2002) noted that R(τ) often becomes noisy at large τ and presented a few methods of circumventing this problem. In Paper 1 of this thesis the integral is truncated at the lowest τ where R(τ) = 0. Using TL we can define the velocity time scale Tv and acceleration time scale Ta (Nilsson et al., 2013; Rupolo, 2007) as
TL + Tv
Ta
=
r
σ2
TL2 − 4 σV2 A
r2 TL −
=
(3.20)
σ2
TL2 − 4 σV2 A
(3.21)
2
Rupolo (2007) calculated both the auto-correlation time scale of the velocity, Tv , as well as the auto-correlation of acceleration, Ta . The ratio of these γR =
Ta , Tv
(3.22)
has been denoted the “Rupolo ratio” where a high γR generally indicates a highly turbulent flow (Nilsson et al., 2013). The results in Paper 1 showed that this ratio generally is very low for the Baltic Sea.
3.4
Lagrangian stream functions
The concept of Lagrangian stream functions was introduced by Blanke et al. (1999). The principle is that any Lagrangian particle trajectory is a transport of mass from one point to another. Similarly to the Eulerian case, a meridional overturning stream function can then be defined for the atmosphere as ∂ Ψ(y, p) = ∂p
I x
g−1 v dx,
∂ Ψ(y, p) =− ∂y
I x
g−1 ω dx,
(3.23)
where v and ω are the meridional and vertical velocities of the Lagrangian particle at latitude y and pressure p. If there are more than one Lagrangian trajectory the velocities can be gridded onto a three-dimensional field and a Lagrangian stream function representing all particles can be calculated using eq. (3.23). In this case a stream line in the Lagrangian stream function will represent the particle path averaged over several particles. A Lagrangian stream function can be calculated for a selection of particle trajectories. By sorting them into different classes and calculating the Lagrangian stream 27
function for each class the total stream function can be decomposed into different kinds of motion. Decomposition of Lagrangian stream functions similar to those in eq. (3.23) were used by Jönsson et al. (2011) in the Baltic Sea and by Döös et al. (2008) in the Southern Ocean to understand the transports of different water masses. The method was applied to the atmosphere in Paper 2 of this thesis to study the transport of air masses to and from the tropics.
28
4. Transports in the Baltic Sea and environmental risks
The Baltic Sea is a relatively small sea enclosed by a large population. The only connection to the world oceans are the narrow Danish straits which makes the Baltic Sea particularly sensitive to pollution. Shipping and near-shore activities can thus pose an environmental risk for the life in and around the Baltic Sea. Although oil spills are uncommon, oil tankers carrying 100000-150000 tonnes1 of oil are continuously crossing the sea (HELCOM, 2010). In the case of an accident, Lagrangian trajectories provide a valuable tool to model the transport of pollutants. An example was given in Fig. 1.1 showing the forecasted destinations of oil slicks in the Gulf of Mexico. In the Baltic Sea, forecasts of Lagrangian particles are produced by the Swedish Meteorological and Hydrological Institute (SMHI) using a particle tracking model and the HIgh Resolution Operational Model for the Baltic sea (HIROMB) model (Funkquist, 2001). Naturally, Lagrangian trajectories can also be calculated backward in time to find where the pollution came from. HIROMB is an operational model that continuously produces 48-hour forecasts. For studying the circulation of the Baltic Sea it is however more common to use ocean models that run for a considerably longer time. One such model is the Rossby Centre regional Ocean climate (RCO) model (Meier et al., 1999) which has been used to produce hindcasts as far back as the 1960’s (Meier et al., 2003). Data from the RCO model and TRACMASS trajectories were used by Döös et al. (2004) to study the large-scale circulation of the Baltic Sea and by Jönsson et al. (2004) and Jönsson et al. (2011) to understand the transport in and out of the Gdansk Bay and the Gulf of Finland. Corell et al. (2012) used the RCO model with TRACMASS trajectories to study the transport and spreading of larvae and how Marine Protected Areas (MPAs) are connected. Soomere et al. (2010) and Soomere et al. (2011) used similar methods to calculate the transport of future oil spills in the Gulf of Finland and used that information to find the shipping routes that would represent the smallest environmental risk in the event of an accident. In Paper 1 we find that TRACMASS trajectories driven by RCO model output have less absolute and relative dispersion (eqs. (3.16) and (3.17)) than the trajectories of observed SVP drifters. This implies that the currents in the RCO model are slower and less variable than those observed, which is a result that will have impacts on future transport modelling in the Baltic Sea. 11
tonne = 103 kg
29
30
5. The overturning circulation of the atmosphere
5.1
Past and present
One of the key features of the atmospheric circulation is the way air masses are transported between the warm tropics and the cold polar regions. Fig. 5.1 shows three of the early conceptual models of the equator-to-pole flows in the atmosphere. Halley (1686) (finder of “Halley’s” comet) and Hadley (1735) proposed that warm, moist air can rise in the tropics because it is lighter than the surrounding air. As it rises the moisture condenses, releases latent heat and forms clouds and rain. At high altitudes, air masses flow towards the poles and radiates some of their energy to space. Hence, near the poles the air sinks as it becomes cold and dense. The loop is then closed by air near the surface flowing towards the equator. In this “Halley-Hadley” model the atmosphere acts as a heat engine that transports air masses from the warm tropics to the cold polar regions. The Halley-Hadley model as shown in Fig. 5.1a is, however, not correct. The Halley-Hadley model predicts equatorward winds at the surface in the midlatitudes (∼ 60◦ N/S) while measurements found them to be poleward on average. Ferrel (1859) and Thomson (1912) therefore, independently, “corrected” the Halley-Hadley idea by adding a cell in midlatitudes. Thomson (1912) added the cell beneath the “Hadley” cell (Fig. 5.1b) while Ferrel (1859) suggested a three-cell structure in each hemisphere (Fig. 5.1c). The “Ferrel” cell was based on the idea of an imbalance between air pressure and the Earth’s rotation causing surface winds in midlatitudes to be southwesterly on average. The most prominent east–west flow is the Walker circulation, which comprises several cells, of which the most well known is the Walker cell over the tropical Pacific Ocean (cf. Peixoto & Oort (1992)). Air rises over the western Pacific and sinks over the east Pacific. The cell is closed by a westerly flow at high altitude and easterly flow near the ocean surface. It was discovered by Walker (1924) who described a surface air pressure gradient over the tropical Pacific Ocean and noted that the rainfall over the west Pacific increased when the pressure gradient was strong. Bjerknes (1969) realised that the pressure difference was associated with a zonal overturning cell which he named the Walker cell. The Walker cell is linked to the sea-surface temperature (SST) gradient over the tropical Pacific Ocean (Neelin et al., 1998; Philander, 1983; Rasmusson & Carpenter, 1982). Periods of small SST gradients, known as El Niño events, are associated with a weaker Walker cell and periods of large SST gradients, 31
height
a) Halley-Hadley model H
90°S
60°S
H
0°
30°S
30°N
height
90°N
b) Thomson model H
F 90°S
60°S
H
30°S
0°
F
30°N
height
60°N
90°N
c) Ferrel model P
90°S
60°N
F
60°S
30°S
0°
P
F
H
H
30°N
60°N
90°N
Figure 5.1: The equator-to-pole motion of air masses as imagined by a) Halley (1686) and Hadley (1735) and b) Ferrel (1859) and Thomson (1912). “H”, “F”, and “P”, denote “Hadley”, “Ferrel”, and “Polar” cells. The Halley-Hadley model comprises two large “Hadley” cells that span the Northern and Southern hemispheres respectively. The Thomson model includes a shallow “Ferrel” cell in the mid-latitudes, while the Ferrel model has a three-cell structure in each hemisphere.
32
a)
b)
c)
d)
Figure 5.2: Meridional overturning circulation of the atmosphere using various vertical coordinates. Positive values indicate clockwise circulation. Crosses mark the maxima/minima of the stream functions, and dashed lines show the mean surface values. Units in Sverdrup (1 Sv = 109 kg s−1 ). Adapted from Kjellsson et al. (2013).
known as La Niña events, are associated with a more intense Walker cell. Fig. 5.2a shows the meridional overturning stream function in pressure coordinates, Ψ(y, p), (eq. (2.6)) calculated from ERA-Interim reanalysis data covering 1979-2009. The picture shows three distinct cells on each hemisphere, similarly to the picture depicted by Ferrel (1859). The three cells are commonly known as the Hadley, Ferrel and Polar cells (Grotjahn, 1993). Fig. 5.2b shows the meridional overturning stream function calculated using latent heat, l [J kg−1 ], as vertical coordinate, i.e. Ψ(y, l). The latent heat of an air parcel is the heat that would be released if all the water vapour would condense. It can be calculated as l = Lv q [J kg−1 ] where Lv is the latent heat of condensation and q is the specific humidity of the air. The stream function Ψ(y, l) shows that the Hadley cells have an equatorward flow at high l and a poleward flow at low l as also found by Döös & Nilsson (2011). The opposite is true for the Ferrel cells. This indicates that the Hadley cells transport moisture from the subtropics into the tropics, and the Ferrel cells transport moisture from the subtropics into the midlatitudes. Furthermore, the Ferrel cells have a much higher amplitude than the Hadley cells in Ψ(y, l) indicating that they facilitate higher moisture transport. 33
1
2
c)
d)
3
Figure 5.3: The hydrothermal stream function, Ψ(l, s), calculated from ERA-Interim reanalysis 1979-2009. Negative values (blue colours) indicate anti-clockwise circulation. Units are in Sverdrup, 1 Sv = 109 kg s−1 . The dashed line is the 345 kJ kg−1 MSE line. The double-dashed line is the Clausius-Clapeyron relation for saturated near-surface air (eq. (5.1)). From Paper 3 of this thesis with permission from AMS.
Fig. 5.2c shows the meridional overturning stream function where dry static energy is used as vertical coordinate. Dry static energy is the total energy that the air would have if it was at rest (i.e. no wind) and dry (i.e. no humidity). It comprises two parts; heat, c p T , and potential energy, gz. Dry static energy, s = c p T + gz, is approximately conserved for adiabatic processes 1 . The result of Ψ(y, s) is that the Hadley cells and Ferrel cells merge to form one circulation in each hemisphere with an equatorward flow at low s and a poleward flow at high s, similar to e.g. Townsend & Johnson (1985) and Held & Schneider (1999). This implies that the dry static energy transport is poleward at all latitudes. Fig. 5.2d shows the meridional overturning stream function using moist static energy as a vertical coordinate. Moist static energy, h = s + l [J kg−1 ], is the total energy of air at rest 2 . The stream function with moist static energy as a vertical coordinate, Ψ(y, h), shows a single circulation in each hemisphere with poleward flow at high h and equatorward flow at low h, similar to Pauluis et al. (2008) and Döös & Nilsson (2011). Note that the picture is very similar to the Halley-Hadley model in Fig. 5.1a. Fig. 5.2d implies that the moist static energy transport is poleward at all latitudes. As shown by e.g. Döös & Nilsson (2011), the poleward energy transport attains its maximum of ∼ 5 PW in the midlatitudes, in close agreement with estimates by Trenberth & Caron (2001). Paper 3 and 4 of this thesis describe the global atmospheric motions from a thermodynamic perspective by calculating a stream function where latent heat and dry static energy act as coordinates, i.e. Ψ(l, s), as formulated in eq. (2.8). This is denoted the hydrothermal stream function and the result when using ERA-Interim reanalysis data 1979-2009 (same as in Fig. 5.2a-d) is shown in Fig. 5.3. The hydrothermal stream function is the sum of all mass transports in the atmosphere that are associated 1 Dry
static energy, s, is similar to potential temperature, θ , so that s ≈ c p θ . static energy, h, is similar to equivalent potential temperature, θe , so that h ≈ c p θe .
2 Moist
34
2000
RCP4.5 RCP6 RCP8.5 Hist. Pre-ind.
1800 CO2 concentration [ppmv]
1600 1400 1200 1000 800 600 400 200
1800
1900
2000
2100 2200 Year
2300
2400
2500
Figure 5.4: CO2 concentrations for climate-model simulations of the period 1765-2100 taken from Meinshausen et al. (2011). Values are shown for the different CMIP5 experiments, “pre-industrial” (1765-1850), “historical” (1765-2005), “RCP4.5” (1765-2100), “RCP6” (1765-2100) and “RCP8.5” (1765-2100).
with a change in either latent heat or dry static energy. Following the outer stream line in Fig. 5.3 shows that the global atmosphere comprises three distinct processes which are shown as arrows. The flow marked as “1” is a flow where latent heat is converted into dry static energy, which is similar to precipitation as in the rising branch of the Hadley cells (Figs. 5.1 and 5.2). The flow marked as “2” is dry air where the dry static energy decreases, which indicates radiative cooling at high altitude. Lastly, “3” marks a flow where latent heat and dry static energy increase. This could be due to moistening and warming of air near the surface or by mixing with moist air in shallow convective cells. The hydrothermal stream function combines the thermodynamic effects of the Walker circulation, the Hadley circulations and the midlatitude low- and high-pressure systems into a single circulation in l–s space (Kjellsson et al., 2013). It thus provides a convenient measure to study the variability of the global atmospheric circulation.
5.2
Future
Observed CO2 concentrations and those projected by various “emission scenarios” (Meinshausen et al., 2011) of the 21st century are shown in Fig. 5.4. Emissions of greenhouse gases (GHG) and tiny particles known as “aerosols” have been shown to have an effect on the global climate system (IPCC, 2013) by e.g. increasing the surface air temperatures globally (Estrada et al., 2013; Hansen et al., 2005, 2013). The global warming and its effects have been extensively studied using climate models, recently in the Coupled Model Inter-comparison Project phase 5, CMIP5 (Taylor et al., 2012). It has been found that warming at the Earth’s surface will in turn cause changes to e.g. the global transports of mass and energy in the atmosphere and oceans (Caballero 35
present
future
convection
radiative cooling P2 = P1 ⇤ (1 + 0.02 ⇤ T ) M2 = M1 ⇤ (1 0.05 ⇤ T )
P1
s ≈ 310 kJ/kg
M1
moistening
q1
q2 = q1 ⇤ (1 + 0.07 ⇤
surface
T)
surface
Figure 5.5: Schematic description of the thermodynamic processes controlling the intensity of the atmospheric circulation and the change from present-day to the end of the 21st century according to Held & Soden (2006). P is precipitation, M is convective mass flux and q is specific humidity, related by P = Mq. Subscript 1 denotes present and 2 denotes future. s = 310 kJ kg−1 is a dry static energy surface that does not intersect the ground.
& Langen, 2005; Hwang & Frierson, 2010; O’Gorman & Schneider, 2008; Vecchi & Soden, 2007; Weaver et al., 2012). For instance, warming of the surface air temperatures are projected to warm and freshen the surface waters in the North Atlantic, which would slow down the Atlantic Meridional Overturning Circulation (AMOC). In the atmosphere, warmer air can hold more water vapour which will lead to a more moist atmosphere in the global mean. As a result of global warming, global-mean precipitation is projected to increase by about 1 − 2 % for each degree of warming while the atmospheric circulation is projected to slow down (Held & Soden, 2006; Vecchi & Soden, 2007). It is possible to estimate the impact of global warming on the atmospheric circulation using some thermodynamic arguments. The Clausius-Clapeyron (Clapeyron, 1834; Clausius, 1850; Wallace & Hobbs, 2006) relation gives the maximum amount of water vapour that air of a certain temperature can hold. dqs qs Lv = , dT Rv T 2
(5.1)
where qs [kg kg−1 ] is the saturation specific humidity, i.e. the amount of kilograms of water vapour that 1 kg air of temperature T holds when the relative humidity is 100%. Eq. (5.1) can be simplified by letting Lv , the latent heat of vapourisation, be constant and evaluating how qs changes for a given amount of warming. ∆qs Lv = α(T )∆T = ∆T, qs Rv T 2
(5.2)
where T = 273 K yields α ≈ 7 % K−1 . Eq. (5.2) thus means that the global-mean specific humidity, q, will increase with about 7% for each degree of global warming, assuming that relative humidity is unchanged (which Held & Soden (2000) found it to approximately be). Held & Soden (2006) found that the climate models used for 36
the 4th Assessment Report of the IPCC (Solomon et al., 2007) showed an increase in column-integrated q of 7.5% K−1 . They set up a simple scaling argument where the precipitation, P, is equal to the upward flux of moisture which is the product of upward mass flux of air, M, and the specific humidity of that air, q, i.e. P = Mq.
(5.3)
∆P/P = ∆M/M + ∆q/q
(5.4)
Differencing,
The last term is equal to the left hand side of eq. (5.2) under the assumption that relative humidity is constant. Climate-model simulations have shown that precipitation increases by 2−3% per degree of global warming (Allen & Ingram, 2002; Held & Soden, 2006). Inserting numbers into eq. (5.4) thus gives: If precipitation increases by 2% K−1 and specific humidity increases by 7% K−1 , then the upward mass flux must decrease by 5% K−1 . A schematic representation of the processes involved are shown in Fig. 5.5. Hence, global warming could cause the atmospheric overturning circulations to slow down. This has been found to be the case in most climate models used in the IPCC reports (Bony et al., 2013; Vecchi et al., 2006). Knutson & Manabe (1995) also suggested that the atmospheric circulation could slow down since the warming of the upper troposphere (about 15 km height) must balance radiative cooling.
37
38
6. Outlook
The most advanced atmosphere/ocean/climate models continue to increase in complexity as they include more and more of the processes in the climate system. The need for tools to evaluate the differences between the models and the real world will therefore also grow. This thesis has presented a set of new tools for doing just that. In the Baltic Sea, a new regional ocean model is being introduced which has yet to be fully validated against surface-drifter observations. Knowledge of its ability to reproduce the transport and spreading of particles will have impacts for e.g. studies of biogeochemistry as well as operational use. The number of surface-drifter observations is very limited and any additional deployed drifters may prove highly beneficial for evaluating the currents in ocean models. For atmospheric purposes, this thesis has also shown that the atmospheric circulation can be very different in different coupled climate models and reanalysis. It has been shown that the models and reanalysis respond to climate variability such as ENSO as well as global warming in a similar way. However, the differences between the models have not been discussed in any greater detail. Specific model simulations where parameterisations or other conditions are changed one at a time could give better insight. To fully understand the origin of these differences the hydrothermal stream function as presented in Papers 3 and 4 could be included in an atmospheric model as an “online”-diagnostic, i.e. be calculated at run-time. This could highlight the impact of various parameterisations, e.g. entrainment, rain auto conversion, turbulent mixing in the planetary boundary layer, etc. The decomposition of the meridional mass fluxes done in Paper 2 can be extended to include other models and a comparison of future and present-day climates. The meridional transport of sensible and latent heat is projected to increase with global warming (Paper 4), but the question remains to what extent this is due to increased mass fluxes or merely an increase in heat content of the air masses. Furthermore, the dependence on spatial and temporal resolution for the conclusions in Papers 3 and 4 remains an open question.
39
40
Sammanfattning
Denna avhandling presenterar olika metoder för att studera rörelser av luft- och vattenmassor i atmosfären och haven samt den transport av energi detta innebär. Resultat kommer dels från olika typer av observationer och dels från datormodeller av atmosfären, haven, och det kopplade klimatsystemet. Ett övergripande tema i avhandlingen är att dels jämföra datormodeller med varandra, och dels att jämföra dem med observationer av verkligheten. Avhandlingens första hälft använder det “Lagrangeska” synsättet där ett flöde betraktas som summan av flera individuella partiklar i rörelse. En del handlar om att förstå hur partiklar i Östersjön följer strömmarna nära havsytan och hur partiklarna sprider sig. Genom att sjösätta flöten som kan spåras med hjälp av satelliter inhämtas information om strömmarnas hastigheter vilket sedan jämförs med resultat från en datormodell. Resultaten indikerar att datormodeller tenderar att underskatta dels strömmarnas hastigheter och dels hur snabbt partiklar sprids, något som kan ha implikationer för såväl studier av biogeokemiska kretslopp som operationella insatser. Liknande metodik appliceras därefter på atmosfärens storskaliga cirkulation för att studera hur luftmassor rör sig mellan tropikerna och mellanbredderna (ca. 45◦ N/S). Det visas att masstransport mot polerna kan tas ut av masstransport mot ekvatorn så att ett medelvärde endast visar mycket små rörelser. Det föreslås att medelvärdet därför kan variera starkt beroende på hur medelvärdet beräknas. Avhandlingens andra hälft fokuserar på det “Eulerska” synsättet där ett flöde ses som en hastighet som varierar i tid och rum. Mer specifikt används s.k. strömfunktioner och en framförallt en termodynamisk strömfunktion där atmosfärens rörelser analyseras utifrån energin i luftmassorna istället för longitud, latitud och höjd/tryck. Denna strömfunktion sammanfattar atmosfären som en värme- och fuktcykel som beskriver bl. a. nederbörd, strålningsavkylning och omblandning i det planetära gränsskiktet. Cykeln visas ha tydlig koppling till El Niño-fenomenet i Stilla havet samt till global uppvärmning. Kopplingen innebär bl.a. att atmosfärens globala cirkulation blir långsammare men att nederbörden ökar. Denna koppling återfinns i flera olika klimatmodeller vilket indikerar att resultaten är robusta.
References
A LLEN , M.R. & I NGRAM , W.J. (2002). Constraints on future changes in climate and the hydrologic cycle. Nature, 419, 224–232. 37 B JERKNES , J. (1969). Atmospheric Teleconnections from the Equatorial Pacific. Mon. Wea. Rev., 97, 163–172. 31 B LANKE , B. & R AYNAUD , S. (1997). Kinematics of the Pacific Equatorial Undercurrent: An Eulerian and Lagrangian Approach from GCM Results. J. Phys. Oceanogr., 27, 1038–1053. 25 B LANKE , B., A RHAN , M., M ADEC , G. & ROCHE , S. (1999). Warm Water Paths in the Equatorial Atlantic as Diagnosed with a General Circulation Model. J. Phys. Oceanogr., 29, 2753–2768. 27 B ONY, S., B ELLON , G., K LOCKE , D., S HERWOOD , S., F ERMEPIN , S. & D ENVIL , S. (2013). Robust direct effect of carbon dioxide on tropical circulation and regional precipitation. Nature Geoscience, Online. 37 B OWMAN , K.P. & C ARRIE , G.D. (2002). The Mean-Meridional Transport Circulation of the Troposphere in an Idealized GCM. J. Atmos. Sci., 59, 1502–1514. 25 C ABALLERO , R. & L ANGEN , P.L. (2005). The dynamic range of poleward energy transport in an atmospheric general circulation model. Geophys. Res. Lett., 32. 35 C HIPPERFIELD , M. (2006). New version of the TOMCAT/SLIMCAT off-line chemical transport model: Intercomparison of stratospheric tracer experiments. Q. J. R. Meteorol. Soc., 132, 1179–1203. 25 C LAPEYRON , P.É. (1834). Mémoire sur la puissance motrice de la chaleur. Journal de l’École polytechnique, 23, 153–190. 36 C LAUSIUS , R. (1850). Über die bewegende Kraft der Wärme und die Gesetze, welche sich daraus für die Wärmelehre selbst ableiten lassen. Annalen der Physik, 155, 500–524. 36 C ORELL , H., M OKSNES , P.O., E NGQVIST, A., D ÖÖS , K. & J ONSSON , P. (2012). Depth distribution of larvae critically affects their dispersal and the efficiency of marine protected areas. Marine Ecology Progress Series, 467, 29–46. 29 DE
V RIES , P. & D ÖÖS , K. (2001). Calculating Lagrangian Trajectories Using Time-Dependent Velocity Fields. J. Atmos. Ocean Tech., 18, 1092–1101. 25
D ÖÖS , K. (1995). Interocean exchange of water masses. J. Geophys. Res., 100, 13499–13514. 25 D ÖÖS , K. & N ILSSON , J. (2011). Analysis of the Meridional Energy Transport by Atmospheric Overturning Circulations. J. Atmos. Sci., 68, 1806–1820. 20, 33, 34 D ÖÖS , K., M EIER , H.E.M. & D ÖSCHER , R. (2004). The Baltic Haline Conveyor Belt or The Overturning Circulation and Mixing in the Baltic. Ambio, 33, 261–266. 29 D ÖÖS , K., N YCANDER , J. & C OWARD , A.C. (2008). Lagrangian decomposition of the Deacon Cell. J. Geophys. Res., 113. 28 D ÖÖS , K., RUPOLO , V. & B RODEAU , L. (2011). Dispersion of surface drifters and model-simulated trajectories. Ocean Modeling, 39, 301–310. 23 D ÖÖS , K., K JELLSSON , J. & J ÖNSSON , B.F. (2013). TRACMASS - A Lagrangian Trajectory Model. In T. Soomere & E. Quak, eds., Preventive Methods for Coastal Protection, Springer International Publishing. 25 D RAXLER , R. & H ESS , G. (1998). An overview of the HYSPLIT4 modeling system of trajectories, dispersion, and deposition. Aust. Meteor. Mag., 47, 295–308. 24
E STRADA , F., P ERRON , P. & L ÓPEZ -M ARTINEZ , B. (2013). Statistically derived contributions of diverse human influences to twentieth-century temperature changes. Nat. Geoscience, 6, 1050–1055. 35 F ERREL , W. (1859). The motion of fluids and solids relative to the Earth’s surface. Math. Mon., 1, 140–147, 210–216, 300–307, 366–372, 397–406. 31, 32, 33 F UNKQUIST, L. (2001). HIROMB, an operational eddy-resolving model for the Baltic Sea. Bulletin of the Maritime Institute in Gda´nsk, 28, 7–16. 29 G ARRAFFO , Z.D., M ARIANO , A.J., A., G., V ENEZIANI , C. & C HASSIGNET, E.P. (2001). Lagrangian data in a highresolution numerical simulation of the North Atlantic 1 - Comparison with in situ drifter data. J. Mar. Sys., 29, 157–176. 23 G ENT, P.R. & M C W ILLIAMS , J.C. (1990). Isopycnal Mixing in Ocean Circulation Models. J. Phys. Oceanogr., 20, 150– 155. 19 G ROTJAHN , R. (1993). Global Atmospheric Circulations - Observations and Theories. Oxford University Press. 33 H ADLEY, G. (1735). Concerning the Cause of the General Trade-Winds. Philosophical Transactions (1683-1775), 39, 58–62. 31, 32 H ALLEY, E. (1686). An historical account of the Trade Winds, and Monsoons, observable in the seas between the Tropicks, with an attempt to assign the physical cause of the said Winds. Physical Transactions of the Royal Society of London, 58–62. 31, 32 H ANSEN , J., S ATO , M., RUEDY, R., NAZARENKO , L., L ACIS , A., S CHMIDT, G.A., RUSSELL , G., A LEINOV, I., BAUER , M., BAUER , S., B ELL , N., C AIRNS , B., C ANUTO , V., C HANDLER , M., C HENG , Y., D EL G ENIO , A., FALUVEGI , G., F LEMING , E., F RIEND , A., H ALL , T., JACKMAN , C., K ELLEY, M., K IANG , N., KOCH , D., L EAN , J., L ERNER , J., L O , K., M ENON , S., M ILLER , R., M INNIS , P., N OVAKOV, T., O INAS , V., P ERLWITZ , J., P ERL WITZ , J., R IND , D., ROMANOU , A., S HINDELL , D., S TONE , P., S UN , S., TAUSNEV, N., T HRESHER , D., W IELICKI , B., W ONG , T., YAO , M. & Z HANG , S. (2005). Efficacy of climate forcings. Journal of Geophysical Research: Atmospheres, 110. 35 H ANSEN , J., K HARECHA , P., S ATO , M., M ASSON -D ELMOTTE , V., ACKERMAN , F., B EERLING , D.J., H EARTY, P.J., H OEGH -G ULDBERG , O., H SU , S.L., PARMESAN , C., ROCKSTRÖM , J., ROHLING , E.J., S ACHS , J., S MITH , P., S TEFFEN , K., VAN S USTEREN , L., VON S CHUCKMANN , K. & Z ACHOS , J.C. (2013). Assessing “dangerous climate change”: Required reduction of carbon emissions to protect young people, future generations and nature. PLoS ONE, 8. 35 HELCOM (2010). Maritime Activities in the Baltic Sea – An integrated thematic assessment on maritime activities and response to pollution at sea in the Baltic Sea Region. Balt. Sea Environ. Proc., 123. 29 H ELD , I.M. & S CHNEIDER , T. (1999). The Surface Branch of the Zonally Averaged Mass Transport Circulation in the Troposphere. J. Atmos. Sci., 56, 1688–1697. 34 H ELD , I.M. & S ODEN , B.J. (2000). Water Vapor Feedback and Global Warming. Annu. Rev. Energy Environ., 25, 441– 475. 36 H ELD , I.M. & S ODEN , B.J. (2006). Robust Responses of the Hydrological Cycle to Global Warming. J. Clim., 19, 5686–5699. 36, 37 H WANG , Y.T. & F RIERSON , D.M.W. (2010). Increasing atmospheric poleward energy transport with global warming. Geophys. Res. Lett., 37. 36 IPCC (2013). Summary for policymakers. In T. Stocker, D. Qin, G.K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex & P. Midgley, eds., Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. 35 J ÖNSSON , B.F., L UNDBERG , P. & D ÖÖS , K. (2004). Baltic sub-basin turnover times examined using the Rossby centre ocean model. Ambio, 23, 2257–2260. 29 J ÖNSSON , B.F., D ÖÖS , K., M YRBERG , K. & L UNDBERG , P.A. (2011). A Lagrangian-trajectory study of a gradually mixed estuary. Continental Shelf Research, 31, 1811–1817. 28, 29 K ALNAY, E. (2003). Atmospheric Modeling, Data Assimilation and Predictability. Cambridge University Press, Cambridge, United Kingdom. 17 K JELLSSON , J., D ÖÖS , K., L ALIBERTÉ , F.B. & Z IKA , J.D. (2013). The Atmospheric General Circulation in Thermodynamical Coordinates. J. Atmos. Sci.. 21, 33, 35
K LOCKER , A., F ERRARI , R., L AC ASCE , J.H. & M ERRIFIELD , S.T. (2012). Reconciling float-based and tracer-based estimates of lateral diffusivities. J. Mar. Res., 70, 569–602. 26 K NUTSON , T.R. & M ANABE , S. (1995). Time-Mean Response over the Tropical Pacific to Increased CO2 in a Coupled Ocean-Atmopshere Model. J. Clim., 8, 2181–2199. 37 K UHLBRODT, T., G RIESEL , A., M ONTOYA , M., L EVERMANN , A., H OFMANN , M. & R AHMSTORF, S. (2007). On the driving processes of the Atlantic Meridional Overturning Circulation. Reviews of Geophysics, 45, 1–32. 20 L AC ASCE , J. (2008). Statistics from Lagrangian observations. Progress in Oceanography, 77, 1–29. 26 L AC ASCE , J.H. & B OWER , A. (2000). Relative dispersion in the subsurface North Atlantic. J. Mar. Res., 58, 863–894. 26 L ALIBERTÉ , F.B., Z IKA , J.D., M UDRYK , L., K USHNER , P., K JELLSSON , J. & D ÖÖS , K. (2013). Reduced maximum work output of the moist atmospheric heat engine in a changing climate. Manuscript. 21 L UMPKIN , R. & PAZOS , M. (2007). Measuring surface currents with Surface Velocity Program drifters: the instrument, its data, and some recent results. In A. Griffa, A. Kirwan, A. Mariano, T. Özgökmen & T. Rossby, eds., Lagrangian analysis and prediction of coastal and ocean dynamics, Cambridge University Press, New York. 23 L UMPKIN , R., T REGUIER , A.M. & S PEER , K. (2002). Lagrangian Eddy Scales in the Northern Atlantic Ocean. J. Phys. Oceanogr., 32, 2425–2440. 23, 27 M C C LEAN , J.L., P OULAIN , P.M., P ELTON , J. & M ALTRUD , M.E. (2002). Eulerian and Lagrangian Statistics from Surface Drifters and a High-Resolution POP Simulation in the North Atlantic. J. Phys. Oceanogr., 32, 2472–2491. 23 M EIER , H., D ÖSCHER , R., C OWARD , A.C., N YCANDER , J. & D ÖÖS , K. (1999). RCO - Rossby Centre regional Ocean climate model: model description (version 1.0) and first results from the hindcast period 1992/93. SMHI Reports Oceanography, 26. 29 M EIER , H., D ÖSCHER , R. & FAXÉN , T. (2003). A multiprocessor couple ice-ocean model for the Baltic Sea: Application to salt inflow. J. Geophys. Res., 108. 29 M EINSHAUSEN , M., S MITH , S., C ALVIN , K., DANIEL , J., K AINUMA , M., L AMARQUE , J.F., M ATSUMOTO , K., M ONTZKA , S., R APER , S., R IAHI , K., T HOMSON , A., V ELDERS , G. & VAN V UUREN , D. (2011). The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Clim. Chang., 109, 213–241. 35 M OREL , P. & D ESBOIS , M. (1974). Mean 200-mb Circulation in the Southern Hemisphere Deduced from EOLE Balloon Flights. J. Atmos. Sci., 31, 394–408. 13 N EELIN , J.D., BATTISTI , D.S., H IRST, A.C., J IN , F.F., WAKATA , Y., YAMAGATA , T. & Z EBIAK , S.E. (1998). ENSO theory. J. Geophys. Res., 103, 14261–14290. 31 N IILER , P.P., S YBRANDY, A.S., K ENONG , B., P OULAIN , P.M. & B ITTERMAN , D. (1995). Measurements of the waterfollowing capability of holey-sock and TRISTAR drifters. Deep-Sea Res., 42, 1951–1964. 23 N ILSSON , J.A., D ÖÖS , K., RUTI , P.M., A RTALE , V., C OWARD , A.C. & B RODEAU , L. (2013). Observed and Modeled Global Ocean Turbulence Regimes as Deduced from Surface Trajectory Data. J. Phys. Oceanogr., 43, 2249–2269. 27 O’G ORMAN , P.A. & S CHNEIDER , T. (2008). The Hydrological Cycle over a Wide Range of Climates Simulated with an Idealized GCM. J. Clim., 21, 3815–3832. 36 PAULUIS , O., C ZAJA , A. & KORTY, R. (2008). The Global Atmospheric Circulation on Moist Isentropes. Science, 321, 1075–1078. 20, 34 P EIXOTO , J.P. & O ORT, A.H. (1992). Physics of Climate. Springer-Verlag. 31 P HILANDER , S. (1983). El Niño Southern Oscillation phenomena. Nature, 302, 295–301. 31 R ASMUSSON , E.M. & C ARPENTER , T.H. (1982). Variations in Tropical Sea Surface Tempeature and Surface Wind Fields Associated with the Southern Oscillation/El Niño. Mon. Wea. Rev., 110, 354–384. 31 RUPOLO , V. (2007). A Lagrangian-Based Approach for Determining Trajectories Taxonomy and Turbulence Regimes. J. Phys. Oceanogr., 37, 1584–1609. 23, 27 S ALLÉE , J.B., S PEER , K., M ORROW, R. & L UMPKIN , R. (2008). An estimate of Lagrangian eddy statistics and diffusion in the mixed layer of the Southern Ocean. J. Mar. Res., 66, 441–463. 26 S OLOMON , S., Q IN , D., M ANNING , M., C HEN , Z., M ARQUIS , M., AVERYT, K., T IGNOR , M. & M ILLER , H., eds. (2007). Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. 37
S OOMERE , T., V IIKMÄE , B., D ELPECHE , N. & M YRBERG , K. (2010). Towards identification of areas of reduced risk in the gulf of finland, the baltic sea. Proceedings of the Estonian Academy of Sciences, 59, 156–165. 29 S OOMERE , T., D ELPECHE , N., V IIKMÄE , B., Q UAK , E., M EIER , H.E.M. & D ÖÖS , K. (2011). Patterns of currentinduced transport in the surface layer of the Gulf of Finland. Bor. Env. Res., 16A, 49–63. 29 S YBRANDY, A.L., N IILER , P., M ARTIN , C., S CUBA , W., C HARPENTIER , E. & M ELDRUM , D. (2009). Global Drifter Programme Barometer Drifter Design Reference. DBCP Report 4, Scripps Institution of Oceanography, La Jolla. 23 TAYLOR , K., S TOUFFER , R. & M EEHL , G. (2012). An Overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498. 35 T HOMSON , J. (1912). Grand currents of atmospheric circulation. In Collected Papers in Physics and Engineering, 153– 195, British Assoc. Meeting, Dublin. 31, 32 T OWNSEND , R.D. & J OHNSON , D.R. (1985). A Diagnostic Study of the Isentropic Zonally Averaged Mass Circulation during the First GARP Global Experiment. J. Atmos. Sci., 42, 1565–1579. 34 T RENBERTH , K.E. & C ARON , J.M. (2001). Estimates of Meridional Atmosphere and Ocean Heat Transports. J. Clim., 14, 3433–3443. 34 VALLIS , G.K. (2006). Atmospheric and Oceanic Fluid Dynamics. Cambridge University Press, New York. 13 V ECCHI , G.A. & S ODEN , B.J. (2007). Global Warming and the Weakening of the Tropical Circulation. J. Clim., 20, 4316–4340. 36 V ECCHI , G.A., S ODEN , B.J., W ITTENBERG , A.T., H ELD , I.M., L EETMAA , A. & H ARRISON , M.J. (2006). Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature, 441, 73–76. 37 WALKER , G.T. (1924). Correlation in seasonal variations of weather, IX. A further study of world weather. Memoirs of the India Meteorological Department, 24, 275–333. 31 WALLACE , J.M. & H OBBS , P.V. (2006). Atmospheric science: An introductory survey. International Geophysics Series, v. 92, Elsevier, 2nd edn. 36 W EAVER , A.J., S EDLÁCEK , J., E BY, M., A LEXANDER , K., C RESPIN , E., F ICHEFET, T., P HILIPPON -B ERTHIER , G., J OOS , F., K AWAMIYA , M., M ATSUMOTO , K., S TEINACHER , M., TACHIIRI , K., T OKOS , K., YOSHIMORI , M. & Z ICKFIELD , K. (2012). Stability of the Atlantic meridional overturning circulation: A model intercomparison. Geophys. Res. Lett., 39. 36
