Related links: ASTM Standard Practice: Overview of materials  | ASTM notes: Additional information...

Revised method compared to that of John Irwin

Note of November 14, 2003:
This is a set of notes prepared by H.R. Olesen of NERI, Denmark. It is meant only for readers who wish to dive deeply into working with the ASTM procedure. Much of the information found here is not any longer relevant, as the software related to the methodology has changed since I wrote the notes.

My procedure is a bit more complicated than John’s. The reasons are two:

1) I wish to determine modelled off-centerline concentrations for comparison with observations (not just strict centerline concentrations)

2) I want an open system which provides possibilities for other types of analyses than John’s.

In order to achieve this, I have created some extra Fortran programmes, and made my own versions of certain of John’s programmes.

John’s programmes are:

NEWFIT.FOR   (version of February 12, 1999)

ASTM.FOR   (as my point of departure I have used the version of January 7, 1998. John now uses ASTM90.FOR which is newer).

My programmes comprise

NF.FOR (A version of NEWFIT.FOR which has some extra capabilities and which corrects a few errors)

NCCS.FOR (Utility for selecting "Near Centerline Concentrations" from a file with observed tracer concentration data).

AM.FOR (stripped-down version of ASTM.FOR which can create some convenient lists).

CB.FOR (ComBine model results and near-centreline concentrations).

In the following, it will be described in which context the various programmes appear.

The Fortran codes are collected in a Zip file (ASTM_SUP.ZIP). In each of the Fortran codes the first section is a comment section, explaining what the program does, which input it requires, and what output it produces. The Zip file also comprises a small data set, ADD_INFO.DAT, containg additional information concerning Kincaid data (such as conversion factor between ppt and microgrammes pr m3 for each experiment).

Please send an e-mail to me if you download and use the Zip file, as that gives me at chance to communicate future improvements.

Preparation of a data set for model evaluation

The procedure is as follows when we wish to form a data set for model evaluation:

1. Take a data set

Take an experimental data set with a good coverage of samplers along the monitoring arcs.

2. Classification into regimes. This step has to be performed by an expert in the art of model evaluation

Classify the observations into a regime. An ensemble average should be found for each regime. Therefore, the definition of a "regime" is important. Possibility: decide whether the observations in each regime are similar, otherwise redefine the regimes.
The result of the classification is called a 'control' file, CONTROL.TXT. An example of such a file for Kincaid is John's KIZIOL2.TXT.

3. Determine centre-of-mass and s y for each arc (NEWFIT).

Consider an arc, and accept it or reject it for further processing. For instance, it will be rejected if there are unacceptably few monitors.

Determine the center-of-mass and the lateral dispersion, s y, for the arc.
Technically speaking, run NEWFIT with option IPLT=6 resulting in SPLUS.DAT. Also run NEWFIT once more with with option IPLT=8 in order to generate FULLARCS.DAT which is a data set with all arcs, suitable for subsequent input to the program NCCS.

4. Determine average s y for each regime (AM)

Based on all observations in the regime, compute the average sy.
Technically, run AM with SPLUS.DAT as input. AM is a stripped-down version of ASTM. The run takes a long time! (on the order of one hour, depending on PC)
The most important part of the output is the small file SIGLIST.DAT with one line per regime.
AM also produces a listing of the relation between arcs and regimes, REGIMES.DAT.

5. Select "near-centerline" concentrations (NCC’s) in each arc.

Select "near-center-line" concentrations.
Technically, run NCCS which takes its input from both FULLARCS.DAT, REGIMES.DAT and SIGLIST.DAT. The output will be NCCS.DAT. This is a data set with observed values of near-centerline concentrations. It is for use by modellers who can compute concentrations at corresponding points.

The above steps can be performed once and for all for a given experimental data set, provided the set of regimes is defined.

The subsequent steps can be performed by the modellers themselves. In the recipe below, it is assumed that we have a Gaussian model where off-centreline concentrations can be computed from information on centerline concentrations plus information on sigma.

Use of a data set for evaluating a model

6. Run a model

Run a model in order to compute concentrations at all arcs, but along the centerline. Also compute the corresponding s y.

Now, choose between two approaches to analyse the model performance (or try using both approaches):

7a. Combine modelled results with near-centreline concentrations with a variety of analyses in mind.

Combine model results with observed near-centreline concentrations (NCCS).
Technically, run CB (ComBine) with model results (MODEL.KIN) and NCCS.DAT as input. Output is data sets which can be used in analyses with Steve Hanna’s package. When you run CB, the modelled concentration at a given receptor is estimated as correctly as possible (based on the concentration in the plume center line and
s y of the model).

At this step in the procedure, we have data sets ready to be used for analyses of model performance. The data set can be used to diagnose model behaviour - for instance by applying the various tools in the Model Validation Kit.

7b Combine modelled results with near-centreline concentrations with the ASTM methodology in mind

Run the ASTM software with model results and SPLUS.DAT as input. In using this procedure "as is", the modelled concentration at a given receptor is approximated by the concentration in the plume center line. This induces a systematic bias, so a model will seemingly have a tendency to overpredict.


Related links:

ASTM Standard Practice: Overview of materials   | ASTM notes: Additional information... | Model Validation Kit 

Homepage of "Harmonisation..." initiative

This page is maintained by Helge Rørdam Olesen

Document date: March 19, 1999. New pre-ample November 2003.

[Dep. Homepage] Department of Atmospheric Environment, NERI