New Ecmwf Weather Prediction Material

[Iceberg]

Some excellent if indepth material on Numerical Weather predicting.

These are from the ECMWF 2010 courses and they have kindly put all the course material on the site, bless them.

http://www.ecmwf.int/newsevents/training/2010/nwp/index.html

Participants to this course are encouraged to study existing lecture material in preparation for the course. This material can be found on our website on:

http://www.ecmwf.int/newsevents/training/lecture_notes/

http://www.ecmwf.int/newsevents/training/meteorological_presentations/

Overview

The objective of the meteorological training course is to assist Member States in advanced training in the field of numerical weather forecasting.

The course is divided into modules as follows:

NWP-NM 12 - 16 April 2010 Numerical methods, adiabatic formulation of models (5 days)

NWP-PR 19 - 28 April 2010 Predictability, diagnostics and extended-range forecasting (8 days)

NWP-DA 5 - 14 May 2010 Data assimilation and use of satellite data

(8 days)

NWP-PA 17 - 27 May 2010 Parametrization of subgrid physical processes

(9 days)

Numerical Weather Prediction, 2010 2010

NWP Course Notes

NWP Course Presentations

Comp. Course Notes

Getting here

Course contents

Application guidelines

Applications

Practical information

General information for course participants

Lecture Notes

[johnholmes]

Very comprehensive

had a quick look at the adiabatic bit-brought me out in a cold sweat-thank gawd all that lot is behind me.

[J07]

This stuff is not new, it's been there a few years. But it's useful to have this stuff all in one place. The numerical methods is particularly good, summing up all the major schemes like leapfrog and forward time/centred space pretty concisely and methodically! (And without being too dry).

The notes do require reasonable exposure to maths and physics though. The meteorology is almost an aside. This stuff is practically just applied mathematics.

Their section on the spectral method is especially interesting, because it is the method used at ECMWF and also GFS I think. (UKMO is fixed grid, Canadian model uses finite elements...I believe).

When using finite difference techniques for evolutionary problems, we only consider grid-point values of the de-

pendent variables; no assumption is made about how the variables behave between grid points. An alternative ap-

proach is to expand the dependent variables in terms of a finite series of smooth orthogonal functions. The problem

is then reduced to solving a set of ordinary differential equations which determine the behaviour in time of the ex-

pansion coefficients.

That makes the spectral method seem more elegant, though it has limitations and they do list them.

Two key advantages: (avoiding aliasing is a good one)

(ii) Non-linear quadratic terns calculated without aliasing (if computed in spectral space o

quadratic grid).

(iii) For a given accuracy fewer degrees of freedom are required than in a grid-point model.

A major disadvantage (since limited area models are pretty important...):

(v) Spherical harmonics are not suitable for limited-area models.

----

Note that they've only listed the syllabus of the courses, not the material itself. What you can access now is background reading.

Anyway, a good find thanks!