Data Assimilation

1
Data Assimilation

• Janneke de Heij
• Jasper Donker

2
Content

• What is data assimilation
• How does data assimilation work
• Pros and cons of data assimilation
• Conclusion

3
Assimilation

• Assimilate   
• to take in, fit into, or become similar (to)
• Data assimilation The process through which
  observations are united into the models initial
  fields

4
Purpose of data assimilation

• It is a way to improve forecasts for large
  numerical models
• Most applications are in meteorology and
  oceanography

5
The use of previous forecast to create initial
condition

• Information from older observations is retained
• It allows the analysis to combine observations
• It interpolates information between measurements
• Dynamical and physical consistencies
• Numerical consistencies

6
Fundamental assumption

• The short-range forecast is assumed to be good
• This implies that observations are used for small
  corrections
• Large errors between the model and observations
  leads to questioning the observations rather then
  the model
• Good Forecast ? Another Good Forecast

7
Corrections to model forecast
8
(No Transcript)
9
1.First Guess observations

• Using a previous forecast for the first initial
  conditions
• Observations centered around the analysis time
• Data not always available or correct

10
2. Observation increments

• Difference between observation and
• short-range forecast is calculated
• The model is corrected for observations

11
Quality Control On Increment

• For comparison with the good quality first guess
• Unrealistic variations between neighboring
  stations and in time are removed

12
Objective Analysis Procedure

• heart of making analysis
• close resembles the observations or the first
  guess field
• One observation may affect many forecast
  variables

13
Objective Analysis Procedure

• Observation errors
• Instrumental
• Representativeness
• Model forecast errors
• dynamical and physical consistencies
• known typical model error structures

14
3.Analysis increments or corrections

• Summing influences by observations
• Making corrections to the first guess model to
  compensate for the influences

15
4. Analyses

• Starting up new forecast cycle
• Data that is not refuted by observations is
  retained

16
Forecast Model

• integral part of the DA system
• consistent with the model numerics, dynamics, and
  physics
• information from previous observations forward in
  time

17
Next Short-range "Guess

• include more data arriving after the data cutoff
  time
• So the short-range forecast is improved,
• Better analysis for the next cycle

18
Analysis quality

• Analysis is good in normal weather
• Intended to optimize forecast
• There dont has to be an error in your analysis
  when it doesnt match your data
• Problems occur in a period of bad runs

19
Cause of errors

• The first guess is bad
• Good data is rejected
• Data which cant be available in similar
  situations
• Assumptions in the analysis are violated
• Tuning

20
Conclusion

• No recognition of rapidly changing weather
• DA is the best model to predict short range
  forecast
• Takes a lot of computation power

21
You will be assimilated