Data

1
Data Modeling

• Data Collection, Databases, Calibration
  Verification

2
Models Data

• Models without data are guesswork
• Guessing is better than no answer!
• Data collection without models is a waste of time

3
Data Collection Modeling An Iterative Process

• You analyze existing knowledge
• You formulate your model
• Or in science hypothesis
• You analyze important aspect using the model
• You collect data about important aspects
• You evaluate your model and start again

4
The Problem Description Equations

• You have now formulated the problem
• Given the equations
• And made the model
• Now you have to look for data

5
How to Get Data?

• Do not start measuring!
• Look for available data

6
International Organizations

• UNEP
• GEMS,
• IOC
• You can find it through the Internet

http//www.unchs.unon.org/ http//www.cciw.ca/gems
/ http//www.unesco.org/ioc/
7
Scientific Literature

• All kinds of data available
• May be used for other problems than yours

8
Monitoring Programs

• International, national regional programs
• Difficult to access
• Many data of poor quality

SEAWATCH http//www.nrct.go.th/HTMLpages/SW/sw_e
_1.html
9
Remote Sensing

• Data are pouring in
• Many more than normally are used

Chlorophyll from Gulf of Thailand
10
Your Own Measurements

• After thorough analysis of existing data
  information
• And after careful construction of a good model
• (gooddescribing basic behavior of system)
• You may make your own supplementary measurements

11
Intensive Measurement Programs

• Measure in several relevant periods depending on
  the frequency response of the system
• At the same time synchronized measurements
• Use automated equipment

12
What to Do with Data

• Three main things
• 1. Check the quality
• 2. Check the quality
• 3. Check the quality
• Be basically a Nerd

13
How to Check Quality

• You must know what levels to expect
• Compare with the model
• The model may be wrong
• But so may the data

14
Specific checks

• Make descriptive statistics and curve and find
  obvious outliers
• Compare min, max values
• Dependent on time
• Sum checks on fractions
• TNgtNH3NO2NO3
• Mass balances

15
Data Storage Databases

• You want to be able to find your data
• And you want to know how the data are defined
• Use a database Microsoft Access is the best
  choice now

16
Files Spreadsheets

• Only as an intermediate tool
• After a while you can not find the data yourself
• And what about your colleagues
• And when you find them you do not know the
  definitions. E.g. units

17
Do You Want Access to Access?

• Databases Environmental Data
• Relational database
• Construction of Database

18
Tables
Sample ID Variable ID Value 300501-04-96 0755
.25 FLU 1.2 300501-04-96 0755
.25 OO 12.3 300501-04-96 0755
.25 SALI 20.9 300501-04-96 0755
.25 TEMP 1.51 300501-04-96 0755
1.14 FLU 1.3 300501-04-96 0755
1.14 OO 12.4 300501-04-96 0755
1.14 SALI 20.9 300501-04-96 0755 1.14 TEMP 1.5
19
Table Design

• Field Names
• Data Types
• Key

20
Relations

• One to Many
• Several Samples pr. Station
• One to One
• Stations Secrets
• Many to Many
• Sampling Program Variables

21
Why Relational

• To be able to identify data
• By putting all data identifiers in tables
• To save storage area
• And save search time

22
Queries

• Flat File Table
• Search
• Calculations

23
Forms

• Data Input
• Graphing

24
Reports

• For printing of
• Tables
• And Graphs
• In high quantities

25
Macros Modules

• For advanced users
• Which you will soon be
• Can do whatever you want

26
Model Calibration

• The use of data for getting the values of your
  parameters
• The level of the parameter has to be known
• But fine tuning can be done be calibration
• To get a system dependent value of the parameter
• E.g. µ for algal growth, Pmax for photosynthesis

27
Calibration Procedure

• Plot observed and simulated values
• Change parameter until reasonable fit
• Use Least Squares for a more rigorous approach
• Remember local minima!

28
Verification

• Test your prognosis
• You can not prove that a model is right!
• There is an infinite number of models which fits
  a given set of data within statistical acceptable
  limits
• A model can be disproved!

29
Data are Good Thinking is Better

• Do not let lack of data be an excuse for
  inaction or indecision's
• Calculate the costs of no action and relate to
  the cost of waiting for more data

30
The Final Basic QuestionCan Denmark Win on
Friday?