Introduction to Climate Modelling

Ravi S Nanjundiah
Centre For Atmospheric and Oceanic Sciences Indian Institute of Science Bangalore-560012 email:[email protected]

Class Notes for Climate Modelling

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

1 / 22

Outline

Components The Big Picture Currents in the Ocean Land-Atmosphere Interaction Land Processes

How do we model? The Atmosphere Oceans Modelling of Land Sea Ice

The Big Picture Again The Big Picture

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

2 / 22

The Interacting Components

We know that climate at a place is affected by condition of the atmosphere We also know that conditions at the surface viz. if it is on ocean or near a ocean or If it is on a mountain or If it is bare land or a desert or If it is forest covered/grassland etc We need to understand how these different features interact with each other Putting these features together and understanding climate using numerical techniques is climate modelling. What does the big picture look like?

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

3 / 22

The Big Picture

Here we can the various components interacting with each other The energy source is the sun At the surface we could have land or ocean Over land we could have at land/mountains All components interact with each other Perturbations in one could inuence the other Ocean and atmosphere are mainly composed of uids Land surface and sea-ice would be solids

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

4 / 22

Ocean - Atmosphere Interaction

Ocean forms an important component of the climate Its high thermal inertia prevents rapid changes in climate a ywheel.
Temperature in Central India during May can reach 40+ o C at Mumbai rarely goes about 35o C

It is the major source of moisture Circulation in the ocean can play a major role in modulating climate at a place e.g. Gulf Stream and Kuroshio currents. These currents carry warm waters from the tropical latitudes polewards and keep the east coasts of N America and Asia relatively warmer and more habitable. Atmosphere drives ocean through winds. Ocean is the most imporant source of moisture Heats atmosphere through sensible heat and long wave emission. Ocean also acts as carbon sink
Ravi S Nanjundiah (Indian Institute of Science) Climate Modelling . . . Climate Modelling 5 / 22

Ocean Currents

The warm waters from the Gulf Stream go along the East Coast of America Similar is the effect of Kurshio This keeps the coastal region warm Also causes large temperature gradients In mid-latitudes what can intense temperature gradients do? Also note the Labrador cold current What can you expect where the Gulf Stream and Labrador cold current meet?

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

6 / 22

Land-Atmosphere Interaction

Let us examine interaction between land and atmosphere Land has lower thermal inertia than ocean can heat up rapidly than ocean Friction from land would be higher The type of interaction can depend on the type of land forest covered or bare land Heats atmosphere through sensible heat and long wave emission. Albedo of the land (dependent on land cover) can signicantly affect climate Land cover can also affect the moisture exchange between land and ocean

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

7 / 22

Land Based Processes

On Land we have further interactions related to climate The precipitation (rainfall or snowfall) can runoff forming streams which join togesether to form larger streams/rivers The precipitation can inltrate into deeper ground This inltrated water can either stay or ow as underwater streams These underwater streams may join with surface streams (it is estimated that signicant part of Kaveri in the delta region ows as sub-surface streams). Precipitation falling over higher reaches of tall mountains could form permanent ice-cover (and signicantly modify the albedo) Some of the ice may start moving down as glaciers these in turn could feed into perenial rivers Ganga, Yamuna, Bhramaputra and Indus get signicant part of the ow especially in the lean season from melting of glaciers. River ow discharges into oceans can change the salinity this salinity change can have signicant impact on the upper layers of the ocean and hence on ocean-atmosphere interaction (processes over Northern Bay of Bengal are an example of this). We need to model these interactions (atleast the major ones)

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

8 / 22

How Do We Model Components?

We have identied four components that interact with each other

Atmosphere Ocean Land Sea-ice

We now need to look at each of them in detail and for understanding climate the major processes in the oceans need to We need to identify major processes that affect these components and model them Some processes may be done in an empirical fashion or the actual process may be replaced by an abstraction of the effects on the system cloud processes are replaced by a parameterization. We now examine the individual components

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

9 / 22

The Atmosphere

We begin with atmosphere we are most familiar with this. Air is a uid We have already identied equations that govern movement of uids
Equation of Mass conservation Equation of Momentum Conservation Equation of State Equation of Conservation of Energy Equations for Conservation of species concentration

Since climate models will generally have coarse grids we may need to approximate some of the processes involved in these equations These equations would be solved for long periods of time we would need very stable numerical methods to solve them

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

10 / 22

Modelling of Oceans
Water is also a uid. Hence equations similar to atmosphere can be applied Circulation in the ocean could be driven either by winds (wind-driven ocean circulation) or By temperature and salinity gradients (thermo-haline circulation). Here salinity of the ocean plays a major role in determining the circulation However we can make some simplifying assumptions to the ocean Density of water does not change much hence we can approximate the equation of state with an empirical relationship as a function of temperature and salinity Unlike atmosphere that is global in nature, oceans could be enclosed/ semi-enclosed we need to consider rigid impermiable boundaries. Flow from rivers/precipitation from atmosphere need to be considered while determing salinity Bay of Bengal and West Pacic (where large precipitation causes a fresh water lens) are examples where this plays an important role. At higher latitudes, ice can form or ice oating down can melt. This also needs to be considered. High thermal inertia causes diurnal cycle to be small. However in regions where the mixed layer is thin, diurnal and intraseasonal variations could be signicant.

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

11 / 22

Modelling of Processes on Land

The land surface is varied : it could be soil, vegetation, snow, glaciers, inland water, mountains and urban areas. Land interacts directly with atmosphere above. With atmosphere land exchanges momentum, energy and mass (water, aerosols and other constituents). Land and ocean are linked through river runoff and breezes due to surface inhomogeneity (land-sea contrast). Unlike oceans its thermal inertia is small. We can also neglect horizontal transport of heat over land Temperatures here have large diurnal cycle. Land cover changes can also cause change in climate (de-forestation). Land acts as a sink for carbon (in land ecosystems).

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

12 / 22

Land Surface Processes . . .

Precipitation from the atmosphere is the major source of moisture The precipitation could either fall directly to the soil or intercepted by canopy (vegetation) The fallen precipitation could either inltrate or runoff The type of vegetation could inuence evapo-transpiration It also reects solar radiation and emits long-wave radiation Various chemical constituents such as CO2 , N2 O, CH4 and VOCs are emitted to the atmosphere. Land surface processes are closely linked to the atmospheric boundary layer There are models of varying degree of complexity The simplest is model representing hydrology in the form of a bucket. More complex models could include some of the processes shown above such as the impact of vegetation and emission of other chemical constituents. When we look at surface processes we also need to estimate the temperature of the surface and near-surface air coupled to the hydological/biological effects. Let us examine a simple model for surface processes
Ravi S Nanjundiah (Indian Institute of Science) Climate Modelling . . . Climate Modelling 13 / 22

A Simple Model for Surface Processes

Figure: A Simple model with no ground inltration

Figure: A Simple model with ground inltration

Till the bucket is full, no water runs off dW = P E W < Wfc dt We can have a tilted bucket dW = P E R if W < Wfc dt R is runoff and could depend on the saturation of soil. In a bucket model with hole R = Rs + Rinl Ground temperature is calculated by solving the Fourier Conduction equation with sensible heat, latent heat and radiation exchanges as the upper boundary condition
Ravi S Nanjundiah (Indian Institute of Science) Climate Modelling . . . Climate Modelling 14 / 22

A More Complex Model : BATS

This is a much more complex model Considers multiple layers of soil Rain is seperated into (a) falling onto bare ground (b) onto vegetation Rain fallen on vegetation in turn could fall on ground as leaf drip We need to consider transpiration from the vegetation and evaporation from the soil seperately

Size of leaves may play a role in interception of rain and also in transpiration Resistence of the stomata to evapo-transpiration could be a function of sunlight, temperature and other factors. Radiation no longer interacts with a single surface; we need to consider both ground and vegetation. The soil could interact with ground water in the deeper parts of the ground

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

15 / 22

Modelling Sea Ice

(based on review by Cecilia Bitz of Univ of Washington)

Figure: The ne picture

Figure: The big picture

Sea ice may not be continous but sometimes on larger scales we may assume continuum We may parameterize sea ice thickness on the sub-grid scale When we look at movement of sea ice we do not consider every single ice-oe but for the continuum (just as in case of clouds) The governing equations for sea ice are Equation for Ice Thickness distribution Equation for conservation of momentum Equation of Enthalphy (energy required to melt sea ice) Heat equation in ice and snow
Ravi S Nanjundiah (Indian Institute of Science) Climate Modelling . . . Climate Modelling 16 / 22

Governing Equations in Sea Ice Model

The rst equation is for Ice Thickness Distribution Dg = g u + (fg ) + Dt h g (h) is a probability distribution function of ice-thickness that the probabiltiy of ice cover thickness h is g , f represents the net growth rate The rst term on rhs represents change in g due to parcel convergence The second term represents mechanical redistribution (essentially deformation) Third term represents change due to growth/melt. Term 4 on rhs is changes due to lateral melt Some simpler models take a single thickness for the entire grid and do not use probability distribution. They will have an equation for gridcell mean ice

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

17 / 22

Governing Equations in Sea Ice Model . . .

The second equation is for conservation of momentum m Du = mf k u + a + w mgr Y + Dt

LHS is lagrangian derivative of u following an ice parcel First term on RHS is the coriolis term, terms 3 and 4 are air and water stresses (water stress will be below the ice). Term 5 is effect of gravity. Y is sea surface height gr acceleration due to gravity. The last term represents convergence of ice stress ( )

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

18 / 22

Governing Equations in Sea Ice Model . . .

The third equation is for energy required for melting of ice (commonly called as enthalphy E) DE = E u + + Dt
The LHS is rate of change of E following an ice parcel First term on rhs is rate of change of E due to convergence Second term is mechanical redistribution of E Last term is change of E due to thermodynamics

The fourth Equation is the heat equation in ice and snow: c T T = k + Qsw (z ) t z z

The LHS is thermal energy change in a layer (at some location) First term on rhs is gradient of conductive ux and second term is radiation term in Enthalphy is partly related to heat equation Growth and Melt also inuence (related to brine pockets inside an ice-oe).

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

19 / 22

The Big Picture Again

We have had a birds eye view of the components of a climate model We have four interacting parts representing four major interacting components These need to co-evolve in an interactive fashion

For atmosphere and ocean we have a set of non-linear partial differntial equation (later when using Semi-Lagrangian techniques well see that we could have coupled ODEs) Need to be discretized and solved Extremely computationally intensive, typically needs 100s of hours of CPU on massively parallel computers. We extensively use parallel computing techniques for this purpose Efcient use of computational techniques as important as other parts of climate modelling
Ravi S Nanjundiah (Indian Institute of Science) Climate Modelling . . . Climate Modelling 20 / 22

Some Computers Used in Climate Studies

This computer (Titan at ORNL) has about 10000 compute nodes (think of 10000+ PCs put together) Each blade has 2 two nodes - each node holds 16 core Opteron 6274 CPU and a Tesla K20 GPU Peak speed of 27 Peta Flops Currently the fastest computer (nd out the details at www.top500.org) Total on-board memory is 710 TB Disk storage 13 PB Uses 12.7 MW of power
Ravi S Nanjundiah (Indian Institute of Science) Climate Modelling . . . Climate Modelling 21 / 22

Indias Fastest Supercomputer

It is at CMMACS Bangalore Has 17344 on 1084 nodes Presently around 85th fastest computer in the world Has a peak speed of 300 T Flops (compare with Titan) Is also being used for climate simulations

Ravi S Nanjundiah (Indian Institute of Science)

Climate Modelling . . .

Climate Modelling

22 / 22