Title: Quo Vadit Visual Weather?
1Quo Vadit Visual Weather?
By Jozef Matula
2What is IBL Visual Weather?
- Meteorological Workstation SW providing
- Met. data processing and visualisation
- Interactive forecasting tools
- Forecast production and workflow management
- Batch production
- Extensibility with Python API
- Web services (WMS, WCS, WFS, JMBL)
- Everything in one box or Client-Server
- Highly configurable
- 1.2M lines of C, 120k lines of Python, 40k
lines of XML, 120k lines of Fortran, 220k lines
of configuration
3past YEAR of VW development
- Developers have worked really hard during
43, 2, 1, start!
- 3 maintained development branches
- 3.1.x stable
- 3.2.y stabilising
- 3.3.z development
- Most significant development areas
- Forecast production and workflow
- Grid dataset concept
- Field Diagnostics and Modification (UK MetMorph)
- Batch production improvements
- Tropical cyclone stuff
- Web Services
5FORECAST production
6Forecast Production - Features
- We refactored our editor of features (mainly for
SIGWX requirements) - Every feature is timestamped and geolocated
- Stored and shared in the Feature Database
- Usability enhancements
- Single click drawing
- Curve optimalisation(to reduce noise from
tablet) - Cooperation withunderlying layers
- Work on WFS-T access in progress
7Example of a Feature Product
With courtesy of 21 OWS USAFE, Sembach
8Forecast Production Text forecasts
- XML Message Editor Templates (declarative
language) - User input validation
- Preview before issue
- Internal production and delivery chain
- Different outputs from one form
- Single click for whole production and delivery
- Revision management
- Corrections, amendments and cancelations
- Access to history of issued forms
- Initialisation of form contents from
- VW Mathematical Kernel (Observation, NWP data)
- Plugins in Python
9Forecast Production - Message Editor 2
PDF
AFTN/GTS
10Message Editor Template Language
- XML description language, consisting of
- Model hierarchical data structure definition
- Layout mapping of model to GUI
11Message Editor Building Blocks
- Instance of data model can be transformed in
multiple ways and delivered to multiple places
Output Filters
Delivery
Text Bulletin
Send to AFTN
Form
ODT w. all reports
Print 2x copies
Aerodrome Warning
PDF for Station 1
FTP to Web
PDF for Station 2
Copy to Archive
PDF for Station 3
Production Pipeline
12Message Editor Embedded Custom UIs
13Forecast Production Workflow Management
- Personal Task Manager provides a role based ToDo
list interconnected with production applications
Access to history of work
Colour of task can be used to indicate type or
priority of the task
Hyperlink
State of work
Workflow actions
14The dataset concept
- Well I am the recent forecast who is more?
15Dataset What for?
- If you modify fields, you might want to see the
previous revisions (On Screen Analysis, Field
Modification) - EPS models require completely new dimension(s) in
grid space navigation (perturbations,
statistics fields, probability fields)
Forecast
Level
Run base time
Dataset
16Datasets EPS Example
Dataset is also a global map variable
17CONSISTENT FIELD Modification (MetMORPH)
- Tell me that this is not a magic
18Field Modification
- We redesigned original simple field modification
module into generalised field modification
framework - 2 available implementations
- Old style single-field modification
- UK Met Office MetMorph (successor of OSFM)
- Output is stored in GRIB and available to
visualisation or batch production - Comparing to UKMO Horace 4 MetMorph
- Facelifted and integrated
- Removed some original limitations
- Access to history of changes
19MetMorph Main Features
- Global and mesoscale model merging
- Dynamical field change
- Editing of wind
- Editing of precipitation (enhancing, reduction,
wind advection) - Merging and advecting precipitation from radar
- Orographic modulation
- Field spatial and time smoothing and
interpolation - Variation of changes in time
- Shape and change-vector morphing
- Variation of change-strenght in time
20Met Morph Dynamical MSLP change
21MetMorph Time Linking
22BATCH PRODUCTION IMPROVEMENTS
- Web server cannot imagine his life without
23Batch production improvements
- Parallel execution of batch jobs
- Ability to have user and role based batch
jobs and limit triggers. - Improved triggers for
- Detection of particular NWP parameter forecast
- Area/station list based monitoring of
observations - Detecting completion of a Feature Product
- New jobs
- Field Diagnostics precalculation of grid fields
algorithms - Field Cutting
- BatchOSA, Batch Message Editor production
24Tropical cyclone forecasting tools
- Also an umbrella is one of
25Tropical Cyclone Forecasting Tools
- Just an experimental internal development both
implemented in Python as a kind of tutorial
sandbox example - TC Storm tracking visualisation (parsing the free
text messages from GTS) - Dvorak helper construction
26Tropical Cyclone Forecasting Tools - Examples
27Web Services
- One day if I want to cook an egg Ill make for it
use of
28Present Web Services in Visual Weather
- Multi threaded with control of maximum load
- Implemented as yet another VW service
- WMS
- Exposes ANY product from visualisation
- Configurable dimensions and layer styles
- WCS
- Exposes any grid field (or derived field) from VW
database, including EPS fields. - WFS
- Access to TAF/METAR reports
- We see a lot of potential here (everything can be
seen as a feature!) - JMBL JMOBS and JMGRID
29OGC WS implementation dilemmas
- The challenge for building web applications is to
find the proper border between server and client,
and to choose the proper protocol - Also there are some technical problems
Compression vs. transparency
Rotated WAFC projections
Depending dimensions
MIME types
Non-geospatial projections
2D time
Date line
Dimension naming
Dynamic capabilities
CRSs vs. GRIB/GeoTIFF
Lack of good clients
Elevation units
Absence of GML schemas
30Visual Weather WMS specialities
- PROJ4 and IBL CRS namespace
- CRSPROJ4projlatlong datumNAD83
- CRSIBLTEPHIGRAM
- Cross-sections
- Do you know CRS1?
- How to pass the place information
- Horizontal route
- DIM_PLACELZIB EPSG43264819 EHAM
- Time crosssection
- DIM_PLACEEPSG43264819
- JPEG2000 (good compression with alpha
transparency) but lack of support in
browsers/Flex the same with GeoTIFF
31Visual Weather WS dimensions
- For WMS naming and dimensions are configurable
- No dimensions at all (latest data use case)
- Single TIME dimension (both for observed or
forecasted data from the recent model run) - DIM_RUNTIME dimension doesnt look to be a
good way (non-orthogonal dimension space) - DIM_RUNDIM_FORECAST
- Optional ELEVATION dimension with configurable
ways how to format the values (850 vs. 850hPa).
32Visual Weather WS dimensions (2)
- For WCS we have only static dimensions
- DIM_RUN, DIM_FORECAST, TIME, ( EPS DIM_MEMBER)
time dimensions overlap - For WFS noone cares about dimensions anymore
its like single table SQL - WCS and especially WFS are vulnerable to DoS
attacks!
33RIA powered by pure VW WMS
341600 other improvements, Changes and Fixes ?
35Thank youfor your attention!Questions?
36Thanks to KNMI and particularly Kees and Helen
for the hospitality!