I would like to apply flood routing methods on a raw discharge-time data set.
It just does not work and I guess that the sharp edges of the data are causing the problem. Is there any hydrological or statistical method that I can make my data set smoother with?
The sharp edges correspond pick discharge values.
Some streams are extremely flashy, responsive to rainfall intensity changes. Not all stream types have access to a floodplain, so their response may also appear flashy. Excessive loading of leaves, debris or sediment can sometimes result in water level pulsing. I had a water level transducer affected by sediment from a storm, and had to get it repaired due to the pulsing of level (pressure). If your storm intensity were to be fairly constant with these hydrographs, then it would be time to question why the abrupt changes, and the rapid recession of the blue line after a storm suggests almost no temporary storage -- extremely ephemeral, and the recession slope is greatest at end of storm, as compared to slower recession during storm relative to the others.
If the data is accurate, smoothing it may look prettier to some, but will just damage its ability to represent these systems. As flashy as they appear, I just wonder just how much flooding is occurring. Manipulating data is fraught with issues for scientific work. If you do it, be very clear what was done and why.
Some streams are extremely flashy, responsive to rainfall intensity changes. Not all stream types have access to a floodplain, so their response may also appear flashy. Excessive loading of leaves, debris or sediment can sometimes result in water level pulsing. I had a water level transducer affected by sediment from a storm, and had to get it repaired due to the pulsing of level (pressure). If your storm intensity were to be fairly constant with these hydrographs, then it would be time to question why the abrupt changes, and the rapid recession of the blue line after a storm suggests almost no temporary storage -- extremely ephemeral, and the recession slope is greatest at end of storm, as compared to slower recession during storm relative to the others.
If the data is accurate, smoothing it may look prettier to some, but will just damage its ability to represent these systems. As flashy as they appear, I just wonder just how much flooding is occurring. Manipulating data is fraught with issues for scientific work. If you do it, be very clear what was done and why.
Hi Pooya,
You can apply moving average smoother, polynomial and periodic regression smoothers, kernel smoothing, nearest neighbour and locally weighted regression, and smoothing splines. If you are not familiar with the above methods; depending on your purpose, I advise using simple moving average smoother or nearest neighbour smoothing by keeping the number of neighbours small to prevent distortion of data to a degree.
All the best,
Hakan
It is difficult to comment given the lack of detail in the question. I assume that the figure given is three gauges located at different points in the catchment, and presumably nested with the grey being the smallest, the blue being in the middle, and the orange being the furthest downstream? What are the catchment areas? If there is a significant difference in the catchment areas, then the spatial distribution of rainfall becomes important and you can't just use flood routing alone.
There are lots of ways to filter data, but note that as William says, you need to be careful with manipulating data. Since you are looking at flood routing, making the data smoother will remove the information that is most important.
- Badges
- Science topic
- Similar topics
- Geoscience
- Hydrology