Various experiments have been performed in this study, the results of each experiment can be found here:

• Twin Experiment 1
• Twin Experiment 2
• Twin Experiment 3
• Twin Experiment 4
• Twin Experiment 4 revised
• Twin Experiment 5
• Twin Experiment 6
• Real World Experiment 1
• Real World Experiment 2
• Real World Experiment 3
• Real World Experiment 4

Prior to the twin experiments and the real world experiments an initial test (Test 1) has been performed with the data assimilation settings as described in the section below.

Test 1

Data assimilation settings:

As the first test, the current operational setting is used:

assimilation locations: 5004A10 and 5004A35 (highligted on the figure below)

assimilated parameters: water temperature and algal diatom on both locations

source of uncertainties: algal diatom at some locations (tributaries) and global radiation stochModelConfig.xml

The experiments are performed using FEWS-NIER, which makes use of OpenDA as a data assimilation tool. In this test, the data assimilation experiments are performed over the year 2015. We focus here on the hindcasts. The hindcasts over the year are splitted into daily runs, each run starts at 00:00 (GMT+9).

In this test, an ensemble of size 8 is used in both the EnKF and Dud-EnKF. With this setting, it took the EnKF about 28 hours and Dud-EnKF 75 hours to finish a one year simulation.

Results

EnKF

Figure A.1: Hindcast output at assimilation station 5004A10

Figure A.2: Hindcast output at assimilation station 5004A35

Figure A.3: Hindcast output at validation station 5004A20

Figure A.4: Hindcast output at validation station 5006A35

Dud-EnKF

Figure B.1: Hindcast output at assimilation station 5004A10

Figure B.2: Hindcast output at assimilation station 5004A35

Figure B.3: Hindcast output at validation station 5004A20

Figure B.4: Hindcast output at validation station 5006A35

Figure Draft Longitudinal Profile Water temperature Yeongsan River

PO4P