Sensitivity of atmospheric near-land temperature in Europe to SST - PowerPoint Presentation

Slide1

Sensitivity of atmospheric near-land temperature in Europe to SST

Andrey Vlasenko, Armin Köhl, Detlef Stammer

Institut für

Meerskunde

Universität Hamburg

Hamburg

TASK

Task 1.2.1 Identification of the atmospheric response to ocean surface state changesThe EU FP7 THOR adjoint assimilation system will be used to identify sensitivities of predictable elements over northern Europe, such as air temperature or precipitation, on parameters in the North Atlantic and the Arctic, such as SST, sea surface salinity or sea ice thickness and concentration. The sensitivity information will be used subsequently to unravel the processes affecting the important climate parameters over northern Europe and underlying time scalesTask leading to deliverable D18.

Plan of the Experiment

Spin up the climate model CESAM, until the proper climatology is establishedDevelop such cost functional that measures the

atmospheric near-land temperature in Europe

g

ives the gradient with minimum numerical noise during the adjoin computations

Using TAF AD tool, obtain the

adjoint

of CESAM

Develop a set of filters that remove numerical noise without spoiling the result

The Model

CESAM = PLASIM(atmosphere) + MITgcm(ocean)

The Driver Program

CESAM

Initialize Atmosphere and Ocean

T

he Main Loop

Postprocessiong

Interpolator

RESULT

MITgcm

ocean(1 step)

PLASIM

Atmosphere (10 steps)

Interpolator

The Experiment

SETUP: Grid resolution

In atmosphere is T21 with 10 vertical layers

i

n ocean is

with 15 vertical layers.

Time step: in ocean is equal to 8 hours in atmosphere is equal to 48 minutes.

TASK:

Compute

Cost Functional:

Where ,

is temperature,

is spatial coordinate,

time

steps,

is a point in the middle of Europe, has a value that temperature

values outside Europe have negligible impact in .Adjoint

:The gradient of

𝐽 (adjoint of CESAM) was generated by a special algorithmic differentiation tool TAF.

Problem 1

The unstable mode.

Problem 2 Distribution of the interpolation error, appearing during coupling

Problem 3

Distribution of the values of adjoit of SST represented as histogram

Solutions

Implementation of a cascade filtering: Low-pass filter. An unstable mode, resulting in exponential increase of high frequency, appears during estimation of adjoints. This mode can be removed without affecting the data by applying low pass filter. Grid noise removal filter. The noise appears due to truncation/approximation errors in the coupling routine. The pattern of these errors are almost constant in space and time

and therefore can be easily recognized and subtracted from the data.

Histogram filter.

As above, due to errors on the coupling stage a couple of outliers appear near the sharp boundaries of the continents. The magnitude of outliers are several order bigger than the magnitude of the data. Therefore, they can be easily designated and removed by a histogram filter.

Results

Sensitivity of near surface atmospheric temperature on 15-th of February in Northern Europe to SST 16 hours before the end of the target period.

Results

Sensitivity of near surface atmospheric temperature on 15-th of February in Northern Europe to SST 48 hours before the end of the target period.

Comparison with the results of A. Czasa and C

Frankignoul*

SST regression maps showing the

tripole

(in K, gray shading, dashed contours for negative) and the North Atlantic

horseshoe

patterns (thick contours, every 0.1 K, dashed for negative

).

* A.

Czasa

and C

Frankignoul: Observed Impact of Atlantic SST Anomalies on the North Atlantic Oscillation . J. Cli. (15) 2002.

Results

Sensitivity of near surface atmospheric temperature on 15-th of February in Northern Europe to SST 150 hours before the end of the target period.

Conclusions

1. The sensitivity of atmosphere to SST in a framework of Coupled model can be estimated. 2. It was shown that SST affects the atmosphere in Europe on short time scales of about 1 day mainly kinematically via heating or cooling the air temperature above. 3. On longer time scales of about a few days, dynamic effects become more relevant.

4. Optimal

patterns resemble the regression patterns SST/NAO, suggesting that SST most

efficiently

affects the atmospheric circulation by triggering an NAO type response.

Further Development

• To obtain the atmospheric response to SST in a framework of maximum configuration where all processes associated with moisture are included.• To estimate the atmospheric sensitivity to other oceanic and atmospheric state variables particularly geopotential height and surface pressure.

Thank you

The research leading to these results has received funding from the European Union 7th Framework Programme (FP7 2007-2013), under grant agreement n.308299

NACLIM

www.naclim.eu