Example how to proceed during data processing:

advertisement
The last update of the document:23.9. 2009
Brief description of the software package
(AnClim, ProClimDB, LoadData)
 AnClim software
o Input format: TXT files
o Working with one station at a time, but automated processing of many stations is
enabled as well
o Menu is ordered in a sequence (steps) to be taken during data processing: viewing
data, adjusting (transformation), testing distribution, finding outliers, homogeneity
testing (both absolute and relative homogeneity tests), times series analysis,
filtering output data
o Other tools: filling missing values, creating reference series, automation
o Managing the software: Series Controler (form in the right bottom corner: info
about period, length of series, number of missing values, option for switching
between monthly data or seasonal and annual averages/sums/extremes)
o Settings the software: in menu Options / Settings, documentation can be found
under menu Help / Documentation.
o Working with series: plotting graphs, finding outliers, calculating statistical
characteristics, various tests (the most commonly used during time series analysis)
o Working with two series: merging two series (using differences, ratios/log ratios)
 LoadData – application for loading data from database (e.g. Oracle)
o Creating connection into database (via ODBC)
o Specification what to download (stations, elements, periods), various profiles for
individual database tables, help buttons for display info about stations, elements,
o Tools for importing data from TXT, … files
o Adjusting output (cross tables)
o Output to TXT files, Excel, AnClim
 Running LoadData application from the AnClim
o Download wizard – guides through all the steps
o Transformation to files suitable for AnClim automatically
 ProClimDB software:
o Used for processing whole datasets (all stations at a time)
o There are two main input files in the software: Data file (dbf file with all stations
data) and Data_info file (list of stations with their geography etc., needed in case
you need to compare a station with its neighbours)
o How to proceed: select a function, specify files (use right click to select various
functions from context menu), set options, run the function for a whole dataset.
You can drag and drop the files from Explorer (or any file manager).
o User has full control over the processing all the time, a lot of auxiliary output is
created
o First step of processing is getting information about all available stations (menu
Get Info – Create Info File), their period etc., the other step is importing
geography (menu Get Info – Import geography), then you can calculate statistical
characteristics , correlations, reference series, outliers, etc. A lot of tools for
managing dataset is available as well.
Click the header to Sort/Unsort,
Right click to mark the column for
Find/Replace function, etc.
Right click to launch context
menu with further functions
and hints.
Functionality available during Viewing a file
 Two modes of data processing: monthly or daily data (in AnClim and ProClimDB).
 Homogeneity testing in AnClim: after we export candidate and its reference series to
TXT files (using ProClimDB), we can use automation in AnClim – running homogeneity
tests for differences (ratios) between candidate and its reference series for a whole dataset.
o It is recommended to use several tests: e.g. t-test (on differences), Man-WhitneyPettit test (non-parametric test), SNHT (several modifications), Bivariate test,
Vincent test (two-phase linear regression).
o Further it is useful to run the tests for several types of reference series: based on
correlations, distances, regional average (good for temperature but in case of
precipitations we can get only one meaningful type of reference series)
o Testing monthly, seasonal and monthly averages
Remark: in the later version, it possible to launch AnClim from 8-4 function in
ProClimDB software and to run the tests automatically
 Results from homogeneity testing are imported back to ProClimDB (imported to DBF
files) and further processed.
o Numbers of inhomogeneities detections per individual years or groups of years are
calculated (summed).
o Where the inhomogeneity detection using various tests, various reference series
etc. coincides, we can regard such cases as very probable to be inhomogeneous,
then to go to metadata and verify them etc.
o After we decide inhomogeneites, we can adjust them, and in the end to fill missing
values
Usual scheme of data processing during data quality control and homogenization (preparing data for time series analysis)
Individual software and tasks that it can solve:
Download data from
database (e.g. Oracle)
Quality control
Homogenization
(ProClimDB)
(ProClimDB/AnClim)
(LoadData)
1 0 .0
0 .8
8 .0
0 .6
6 .0
0 .4
4 .0
0 .2
2 .0
0 .0
0 .0
- 0 .2
- 0 .4
- 2 .0
- 0 .6
- 4 .0
1950
1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
- 0 .8
- 1 .0
1911
1915
1919
1923
„Technical“ series and grid points calculation
(ProClimDB)
ProClimDB:
Statistical analysis
SPI, …
Extreme value
analysis
Spatial analysis
(connection ProClimDB - ArcView)
Further tools:
(connection ProClimDB - R)
Validation of RCM
outputs
Correction of RCM
outputs
1927
1931
1935
1939
1943
1947
Download