In April this year, Costanza and I went to try and download some data from a water level guage on the Dudh Kosi Phanka and Nha, in front of the Ngozumpa glacier, operated by the Department of Hydrology and Meteorology. Unfortunately, we found that it had been dismantled – looks like someone decided they could make better use of the solar panel and the battery. To be fair, that is likely to be true!

Anyway it made me think about what data was ever collected from this guage and all I could find was  data from 2007, though there may well be more. I don’t know the full history of its installation, maintenance and data download history.

You can see some of the data below but first some information on the sensor and some caveats.

• The sensor is a sonic ranger so it measures a distance to the water level, by using the speed of soun, and the travel time for a sonic pulse to bounce back from the water surface to the sensor.
• This will obviously be noisy as the water surface is choppy and variable, even though its mounted over a backwash pool.
• Readings were saved for 15 minute, 60 minute and daily values – I have not seen the program for this but I assume these are averaged values in order to filter out some of the noise.
• This location captures runoff from the connected lakes and cathcments on the west side of the glacier, as well a the glacier.

First, I wanted to remove some of the outliers as there are some big jumps in the apparent water level in the dataset which are unlikely to be true. So, I made a very crude filter by determining the mean distance to the water level (calculated on all data including those that I think are probably erroneous), which was 2.01m, and removed all data points that deviated from this by more than 1.50m. I chose this threshold just by eyeballing the data and its a pretty generous filter in that it definitely lets through some remaining noise, but I thought thats better that excluding some valid data.

Here is the daily mean water level for the available data bracketed by the daily maximum and minimum in grey. Clearly there are still some values that are most probably errors and they will be affecting the calculated mean values, but you can see the general pattern of peak water level during the core monsoon months which is as expected.

Then I used the 15 minute data to plot the hour of the day when the high and low water occurred. Becasue of limitations in the size of the memory of teh datalogger (which overwrites its data once its full), this covers only the latter part of 2007. I think the plot is kind of interesting as it looks like during the core summer monsoon months of July, August and September that the river level is highest around the middle of the night, whereas outside of these three summer monsoon months the river is highest in the middle of the day.

To be honest I’m not totally sure how to interpret this. Is this because the monsoon rain comes in the afternoon/evening here and by the time it all comes downstream its late into the night? Or does this tell us something about the efficiency of the glacier drainage system varying between these two seasons? I’d need to analyse the precipitation data at the same time to start exploring this, and I do not have that data to hand just now. Watch this space.