Diagnostic output overview¶

The MIKE FM DA module can output 3 different types of diagnostic outputs:

Measurement diagnostic - which relates to a specific measurement
Non-measurement point diagnostic - results for a specific variable and point
Global assimilation statistics

All are read by the FMDAp method read_diagnostic()

In [1]:

import fmdap

1. Measurement diagnostic overview¶

Measurement diagnostic (type 1) comes in two kinds depending on the type measurement they refer to:

point
spatially distributed (e.g. track measurement)

They furthermore behave differently depending on presence of assimilation updates or not.

Measurement diagnostic have the following main data properties:

forecast
result
innovation

If the file contains updates (from assimilation) it will also have properties:

forecast_at_update
analysis
increment

If the file does not have updates, the forecast and result properties will be identical.

Point measurement diagnostic without updates¶

In [ ]:

Point measurement diagnostic with assimilation updates¶

In [2]:

fn = '../tests/testdata/Diagnostics_F16_EnKF.dfs0'
diag = fmdap.read_diagnostic(fn, name="F16")

In [3]:

diag

Out[3]:

<MeasurementPointDiagnostic> F16 (Significant wave height [m])
 Time: 2017-10-27 00:00:00 - 2017-10-29 18:00:00 (397 steps; 347 with updates)
 Ensemble with 10 members
 Model: 7440 values from 1.7218 to 7.3975 with mean 3.9859
 Measurements: 125 values from 1.7495 to 7.2173 with mean 3.9901
 Mean skill: bias=0.1871, rmse=0.4057 ensemble_std=0.1365

In [4]:

diag.type

Out[4]:

<DiagnosticType.Measurement: 1>

In [5]:

diag.has_updates

Out[5]:

True

Track measurement diagnostic with assimilation updates¶

In [6]:

fn = '../tests/testdata/Diagnostics_Altimetry_C2.dfs0'
diag = fmdap.read_diagnostic(fn, name="c2 alti")

In [7]:

diag

Out[7]:

<MeasurementDistributedDiagnostic> c2 alti (Significant wave height [m])
 Time: 2017-10-27 12:50:00 (1 record with update)
 Spatially distributed points with avg 62.0 points per step
 Ensemble with 7 members
 Model: 434 values from 0.0297 to 2.6706 with mean 0.9674
 Measurements: 62 values from 0.3489 to 2.5001 with mean 1.0032
 Mean skill: bias=-1.0368, rmse=1.1841 ensemble_std=0.1531

2. Non-measurement point diagnostic overview¶

Non-measurement diagnostic (type 2) are always point-type. The don't have any measurement information. They behave differently depending on presence of assimilation updates or not.

Non-measurement diagnostic have the following main data properties:

forecast
result

If the file contains updates (from assimilation) it will also have properties:

forecast_at_update
analysis
increment

If the file does not have updates, the forecast and result properties will be identical.

Non-measurement point without updates¶

In [8]:

fn = '../tests/testdata/diagnostics_nonMeas_SSC1.dfs0'
diag = fmdap.read_diagnostic(fn)

In [9]:

diag

Out[9]:

<NonMeasurementPointDiagnostic> diagnostics_nonMeas_SSC1 (Concentration 2 [kg per m pow 3])
 Time: 2016-01-01 02:00:00 - 2016-01-01 11:25:00 (114 steps; 0 with updates)
 Ensemble with 3 members
 Model: 342 values from 0.1810 to 0.6348 with mean 0.3458

In [10]:

diag.type

Out[10]:

<DiagnosticType.NonMeasurementPoint: 2>

In [11]:

diag.has_updates

Out[11]:

False

3. Global assimilation statistics¶

Currently, global assimilation statistics files have very limited support!

In [12]:

fn = '../tests/testdata/Global_stats.dfs0'
diag = fmdap.read_diagnostic(fn)

In [13]:

diag.type

Out[13]:

<DiagnosticType.GlobalAssimilationStatistics: 3>

In [ ]: