I have a network of rain gauges and stream flow measurement, but often due to some issues I have missed rainfall measurement and stream measurement so how can it be find out on daily time scale.
I tried with the linear as well as multiple regression and found not suitable correlation. Even interpolation during rainfall creates problem in stream measurement series.
Can you please suggest me any conventional way or open software to find out the missing value.
I think interpolation is good Technic for this problem but its accuracy is depend upon how we use it. If we do the simple interpolation without considering other physical factors then error may occur but you try to consider the physical setup between the stations, Relation of rainfall figures between those stations in past and then use interpolation.
Dear Prof. Borse
I did the interpolation but for daily discharge it didn't work out. And what do you mean by physical factors that we have to consider.
like if there is a mountain range in between two stations. One station is on wind word side and other is lee word (rain shadow)side then rainfall decrease towards leeward side rapidly. so here our calculation should be wrong... only distance between two station is not sufficient..
Dear Vikram,
Can you use "persistence" to fill in the data? (Use one data point value until another comes along?)
Jack
the main draw back of interpolation Technic is we suppose that the region is homogeneous. which is not true.
Vikram, Pl have a look at this paper. For rainfall, correlation with the adjoining stations is the best way to fill the missing data. Best,
Article Estimation of Columbia River Virgin Flow: 1879 to 1928
Dear Arun
As mountainous catchment have different climatic condition and it varies spatially and even due to some orographic effect I didn't find good correlation with near by station.
For stream modelling can you suggest me some statistical or non statistical model that deal with lest data.
I suggest you first look at the literature (various journals), e.g. Water Resources Research, Groundwater, J. Hydrology, Mathematical Geosciences, Meteorology, various atmospheric sciences journals. This is not exactly a new problem but neither is there a simple solution that fits every problem.
Most likely you will use one of the forms of kriging (which is essentially the same as Radial Basis functions which in turn includes splines as a special case).
As noted above local geographic conditions are an important factor. Regression is not likely to work because it attempts to predict the value of a dependent variable based on the values of one or more independent variables and the choice of those variables is critical. Moreover rainfall is a spatial phenomenon so you have to take spatial location and spatial proximity into account. Rainfall is also a temporal phenomenon but time series alone is not sufficient.
Since you didn't find a good correlation between rainfall stations, any method will give a lot of uncertainty. There are a few similar questions you might want to look at:
https://www.researchgate.net/post/What_is_the_most_effective_way_to_fill_the_missing_hydrological_data
https://www.researchgate.net/post/How_to_calculate_missing_daily_precipitation_data
If you have flow data but missed the rainfall data, you can use the flow data and the model to get an estimate of the rainfall, but this will depend on the model, and you will need to take care when evaluating the model performance due to serial correlation.
This is a sequential data problem.
One of the most effective solutions is the use of recurrent neural network (RNN).
I am studying this problem.
This technique is ideal to fill gaps, you just give the available time-series and the missing time steps will be simulated with a resampling approach that generates realistic rainfall patterns: Article Missing data simulation inside flow rate time-series using m...
Here you can find a tutorial to use the same technique with a free software. The example is on rainfall data, but you can just adapt it using streamflow series: https://bitbucket.org/orianif/g2s_time_series_tutorial/src/master/Let me know if you need further help. Fabio
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping