Transcript MERSEA/MyOcean meetrika (1)

NSO8055 Okeanograafiline prognoos
Jüri Elken [email protected]
Okeanograafiliste mudelite prognoosi osavus
Üldpõhimõtted
MERSEA/MyOcean meetrika
Taylor’i diagramm
MyOcean eelkäija, 2003-2009
The MERSEA ocean monitoring system is envisioned as an operational network that
systematically acquires data (earth observation from satellites, in situ from ocean observing
networks, and surface forcing fields from numerical weather prediction agencies). These data
from diverse sources are combined and merged with numerical ocean circulation models (i.e.
assimilated) to produce best estimates of the actual state of the ocean (nowcasts) and
forecasts. That information is then disseminated to enable various users to develop specific
applications. The project itself has identified areas of Special Focus Experiments on seasonal
weather forecasting, and on ecosystem modelling in regional and shelf seas. Specific
applications on marine safety concern improved wave forecasts, offshore operations, ship
routing, and oil spill drift.
The project work plan is structured on a set of thirteen work packages organized in three main
modules :
observing systems and provision of data ;
the design, development, implementation, integration, evaluation and validation of a coordinated set of monitoring and forecast systems covering the global ocean and the oceans
and seas surrounding Europe ;
the development and demonstration of information products, applications and services, in
partnership with intermediate users.
MERSEA/MyOcean meetrika (1)
Class 1 metrics
“Class1 diagnostics gathers 2-D and 3-D fields …
interpolated on the GODAE grids, and averaged on daily
means … Class 1 metrics are designed for “consistency
assessment” and comparison to climatologies or ocean
pattern described in the literature.”
Name of Metrics
Description
Supporting observation
SST
night-time SST, 1 model layer, 5km Janssen et al climatology
model grid, whole region
L3P-product from SST-TAC, backup:
L4 -product from SST-TAC
SSS
SSS, 1 model layer, 5km model grid, Janssen et al climatology
whole region
Sea Ice
Sea Ice concentration, 5km model grid, Baltic Sea ice charts
whole region
st
st
MERSEA/MyOcean meetrika (2)
Class 2 metrics
“Class2 diagnostics gathers some of the model variables
… along chosen section tracks or at moorings locations”
Name of Metrics
Description
Supporting observation
Temperature
Time series of temperature profiles at Temperature observations from in-situ
locations of BOOS stations and TAC, BOOS network
monitoring stations of the HELCOM
Baltic Monitoring Programme (BMP), see
figure 1,2, complete list in the appendix,
hourly data at several depth levels
Salinity
Time series of salinity profiles at Salinity observations from in-situ TAC,
locations of BOOS stations and BOOS network, HELCOM combine
monitoring stations of the HELCOM data set
Baltic Monitoring Programme (BMP), see
figure 1,2, complete list in the appendix,
hourly data at several depth levels
Sea level
Sea level at tide gauge stations, hourly Sea level observations from in-situ
data
TAC, BOOS network
MERSEA/MyOcean meetrika (3)
Class 3 metrics
Class 3 metrics are similar to class2, but are derived
quantities, such as fluxes and overturning circulations. They
require inline calculation.
Name of Metrics
Description
Transports
Time series of daily mean volume BOOS model, literature
transports across the BOOS sections
Monthly
currents
Supporting observation
mean Monthly mean currents on model grid, literature
variance and persistency of currents
MERSEA/MyOcean meetrika (4)
Class 4 metrics
Class 4 metrics provide statistical comparisons of model against data.
These will be compiled both for the analysis and forecast steps, to give
an objective indication of the accuracy of the system at all lead times.
The focus will be on SST and Sea level, although there will be some
assessment of profile temperature and salinity, and currents for which
the observational data is more limited. For the call 4 metrics we will
compute at least the statistics detailed in appendix B.
Statistical quantities used for Class 2&4 metrics
The mandatory analysis is restricted to simple statistical quantities. Focus
is on those quantities which are necessary to summarise the output of the
statistical analysis in Taylor diagrams:
• Standard deviation
• Correlation coefficient
• Rms difference
This analysis will be extended by the following quantities in order to
separate the total difference between model and observations into
components:
• Mean error/Bias
• Standard deviation of mean error
Taylori diagrammi
käitumine
võnkumiste
muutmise korral
Näide erinevatest X-Y diagrammidest
ja neile vastavatest korrelatsioonidest
Anscombe's quartet
4 identse statistikaga andmehulka
Üks osavusindeks?