EVALUATION OF ORDINARY KRIGING AND ORDINARY COKRIGING RAINFALL INTERPOLATION METHODS USING ELEVATION COVARIATE VARIABLE
The accuracy of various hydrological analyzes depends heavily on accurate estimates of the rainfall spatial distribution. Satellite data is often used as an alternative grid data in areas that do not have plenty of rain gauge stations but still have low data accuracy. Thus a high-resolution grid...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/46684 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The accuracy of various hydrological analyzes depends heavily on accurate
estimates of the rainfall spatial distribution. Satellite data is often used as an
alternative grid data in areas that do not have plenty of rain gauge stations but still
have low data accuracy. Thus a high-resolution grid data will be created from the
rainfall observation using the best interpolation method.
This study began with searching for the best interpolation method between
Ordinary Kriging (OK) and Ordinary Cokriging (OCK) with monthly rainfall data
for 1998-2016. Then an evaluation was made regarding the effect of elevation and
station samples variables on interpolation results. The best interpolation method
was evaluated and implemented for daily rainfall interpolation. GPM and CHIRPS
data are assumed to be the rainfall products that represent surface rainfall and are
used as the comparison data for the best rainfall interpolation.
The results show that OCK method is better than OK with RMSE OCK 18.13 and
RMSE OK 19.02. The elevation covariate variable can reduce the RMSE value.
Variation in station samples shows higher RMSE result when the number of station
samples decreases by 0.1-2. OCK, GPM, and CHIRPS spatial plots do not show the
same pattern for extreme events. This corresponds to the small correlation values
in the range of 0.1-0.4. Whereas OCK has a fairly good monthly correlation with a
range of values of 0.4-0.7. OCK can detect rainfall with monthly POD values
ranging from 0.7-0.9. Whereas OCK made a mistake in detecting rain events with
daily FAR values ranging from 0.5-0.9. Daily and monthly evaluation results
indicate the OCK method has a high POD value, low FAR, and a pretty good
correlation for the monthly. |
|---|