esteng

Here are source data, models and resulting maps of the study Interpolative mapping of mean precipitation in the Baltic countries using landscape characteristics by Kalle Remm, Jaak Jaagus, Agrita Briede, Egidijus Rimkus and Tiiu Kelviste.


Gradual generation of the map depicting the mean annual precipitation in the Baltic countries

1. Similarity-based estimation using software system Constud at interval 8 km. The result is interpolated if the predictions at corners differ less than 5 mm and the similarity to exemplar stations differs less than 10%.
2. Estimation at interval 4 km.
3. Estimation at interval 2 km.
4. Estimation calculated for all pixels.
5. The Baltic Sea masked out.
6. Coastline, frontiers, larger rivers and observation stations added.



Data for download

Constud mapping 245 stations - Access database used for the machine learning in Constud using data from all 245 stations.
Constud training - test different radii - Access database used for the learning in Constud using learning data set (123 stations), and for validation according to the test data set (122 stations).
DM_projects - Data Miner projects for calibrating and validating models of annual, DJF, MAM, JJA and SON amount of precipitation. An explanation on a bug in the code generator of BRT models is added. Data formats: Statsoft Statistica 9 Data Miner and MS Word 2003.
245 DM models - models of data mining methods calibrated on 245 observations.
Learning DM models - models of data mining methods calibrated on 123 learning observations.
Constud similarity maps - similarity maps generated in Constud. .
Constud estimation maps - maps of estimated values of the mean precipiatation generated in Constud.
DM estimation maps - maps of estimated values of the mean precipiatation generated from data mining models.
RDWM maps 245 - reverse distance weighted mean precipiatation in centimeters within 75 km vicinity. Data from all 245 stations.
RDWM maps 123 - reverse distance weighted mean precipiatation in centimeters within 75 km vicinity. Data from all 123 stations.
Landscape data - Corine 2000 land cover and SRTM elevation model in byte format (elevation>254m recorded as 254 m) as used for calculating locational features.
Local landscape variables - numeric codes of local landscape variables in the datatables and variable names.
DM maps VB - Visual Basic project used for generating raster maps from DM models.

Last update 28.08.2014 Webmaster