There might be some, I don't know.  I see karst on you list, that might be difficult.  Garrett et al, 2011 used what they called End Member analysis to separate flow derived from rain, shallow riparian and deeper groundwater that seemed interesting.  when you mention river, large drainage comes to mind.  If with river records, you can define the recession curve during non-rain periods (this may change with growing season and dormant season) you are well on your way.  You need the before storm baseflow to link to the after storm baseflow curve.  We did this by hand in the old days, first extending the post storm recession backwards into the storm, and then pushing then prestorm baseflow forward into the storm to smooth the curve.  I think this is probably mathematically possible, but not attempted myself.  Hopefully you can find some who have.  If I get into my computer, I will try to send you a copy of the Garrett paper.  Each stream, river may have their own recession curve, with similarities among streams or rivers by physiographic area.  My paper on water quality in springs in Missouri may be useful for karst discussion.  On large rivers, 30 minutes is probably OK, but on smaller hydrologic, not so good.  I think USGS typically uses 5 minutes, for better detail, but if there is an outage, it might mean a bit more data, and there is a log passing that temporarily alters water level at gauge, it is better defined as an anomaly.

Thanks for the suggestion..i'm waiting your favourable 'garret et al' paper. 

Thanks.

Look at this paper and the springs of Missouri paper on my list.  if groundwater has a certain concentration, and is present and adjusts directly or inversely in baseflow as it declines, the dilution of that concentration would be an indicator of what part is groundwater dominated, and what part is rain dominated.

SPRINGS OF MISSOURI PAPER

I've already downloaded you pdf.file,

thanks

Dear Adji, Look at this document in attach.

Dear Adji, Look at this document in attach.

Dear Laouacheria,

Thanks a lot , this will be very helpful

Hi

two papers on this topic.

Using recession curve and precipitation data, plus groundwater annual maximum level over the basin

Best

Gil

Article Surface water and groundwater relationships in a tropical ri...

Article Surface/groundwater interactions in the Bani and Nakambe riv...

Is there any way to do in in excel?

the recession coefficient can be calculated with excell of course, you just need the daily flows.

But to use the methodology we present, you also need to have at least some data from wells groundwater levels, to know at what moment you reach the maximum level each year (usually just after the rainfall peak)

Tjahyo Adji

Does anyone has an experience on how to automatically separate of(out the??) base flow of a one year hydrograph?

How to automatically separate base flow on a one year river hydrograph?The discharge record is in every 30 minute.

- - - - - - - - -

William F. Hansen

Added an answer

" I think this is probably mathematically possible, but not attempted

myself.  Hopefully you can find some who have. "

- - - - - - - - - - -

Greetings Tjhoyo Adji,

Hopefully you are still looking for an answer to your question - because there is a really simple mathematical solution to your problem. You see, your data you collected is what is called, "Time Series"data ... which simply means that you are recording measurements (of a variable "Y", in terms of dx(where x = time, plotted on the 'X'-axis).

I know ... looks like "Calculus", but no need to go there. Instead, you're going to need to borrow one of the Electrical Engineer's math tools - specifically the "Auto-Regressive, Integrated, Moving Average" ... known as ARIMA for short.

It's very Simple. It works really great - using "Excel".

So follow these steps exactly - even though some steps are going to sound 'strange to you'.

First - you need a whole lot of data - because you said that you wanted to remove the annual base flow.

I said a whole lot of data - because this method will take your data ... and then "filter" your data out - into two parts. The first part is going to find the 1-year base flow. You then take this resultant work ... and subtract this filtering data - from you initial data.

This ARIMA is a real, functional filter. By filter, in this case you will convert your original data, converting it into two, different parts. The so-called "annual filter part" where were going to set up to "filter out ALL cyclic components "having a time length equal to or greater than 1-year." The second component "shows" all cyclic/frequency components having a time length less than 1-year ... and that's the part that has 'your answer'.

So, now go to Excel - and open up a blank work sheet.

Name it.

In Column "A" write in your data::time as to when you collected this data.

In column "B", - type in most all your hydrologic data.

How much data?

As much as you need. I'm assuming that you would like to see a graph output representing 31 days worth of data and time. (This should cover even Typhoons and Hurricane events.) If it does not, then lengthen or shorten the graph period of interest ...

but we're going forward now with 31-days for this example.

Now, this ARIMA (i.e., annual Filter), (starting at day - 0"), it needs the earlier(previous) 183 days of data. (Mathematically, this is determined by you're filter length ... which for us is ("n + 1 days")/2 ... or (365 + 1)/2 = 183 days of previous data.

You will then need to determine how much data you need on the tail end.

Same formula: Go to day 31 of the month long series of interest. Then add in the following 183 days of data. ... or (365 +1)/2.

- - - - - -

Since you have data being collected every 30 minutes ... you'll need

(365 + 1 + 31) days of data = 366 + 31 = 397 days worth.

But because you will be using 30 minute data streaming - this becomes

(397-days X 24-hours/day X 2-data points/hr. = 19056 points of hydrology data. !!!

That's a lot.

Now go to Column "C". (I'm a bit rusty on my Excel) so do this.

Go down Column A until you reach the very first data point "of the month of interest.

If you start at cell one, this should be cell (183-day X 24-hours/day X 2 readings/hour) = line #8784.

Now within Cell C8784 write an equation - so that one sums up all hydrologic data collected all the way from line # 1 down to line # 17568. Then divide the resultant, by 17,568.

((365 +1)X24-hrs/day X 2 readings/hr ) = 17568

I'm thinking the equation should look something like this: =Sum(B1:B17568)/17568.

As you should be able to tell, this should be giving you the average value of Hydrologic data - for exactly 1-year ... but note a very important (highly unusual point!) you are plotting this point you just computed exactly half way between the start and the finish points ... That's why you come half way back, to cell 8784.

Now, a simple step. Just grab and copy this above formula down to cell

#10,272. That's (183+31 days) X 24-hrs/day X 2-reading/hr. = 10,272

- - - - - --

Goof news. The hard part is done:

Now go down column D - down to line 8784.

In this cell "D8784" you're supposed to subtract cell C8784 from cell B8784. =B8784-C8784

Thus, you are left with just the difference - of the measured value of the Hydrologic value subtracting the annual moving average rate ... for that time measurement.

You're done! All you have to do is copy down this formula to cell D10,272.

- - - - - - - - - - - --

Now you graph it. Graph the results of columns B, C and D separately - using lines 8784 down to line 10272 ... the 31 days of interest.

Now, I would also go ahead & graph columns B & C on the same graph ... which shows your hydrologic data vs. the annual rate.

Then look at graph of Column D data. That is the 'exact answer' to your question you asked above.

= = = = = = = = = = = = = = = = = = = = = = = =

Now I'm going to show you another simpler example of what is going on. Lets plot

a straight line, equation of "Y" = 21 - X.

This is a straight line. Y = 21, when X = 0

Y = 20, when X =1

Y = 19, when X =2

Looks sort of like this:

*

*

*

*

*

*

*

*

*

*

*

*

^

Now let's just say that this is monthly data - starting with (0,21) being Jan 1'st.

Now watch carefully, and do your own counting. If you go out 12 count - you're only at Dec 1'st, not Jan 1'st. That's 'why' you go (N + 1)-counts.

or one extra to get to Jan 1'st ... in order to pick up the full, entire range of time data.

Now add up the "Y" values of the first 13 months: (21 + 20 + 19 + 18 + 17 + 16 + 15 + 14 + 13 + 12 + 11 + 10 + 9)

= 195

and divide by 13.

= 15

Now go ahead and plot this value on the above graph. (Most people will plot a so-called moving average value by going out 1-more unit of time ... to 13, and go up to plot the above moving average value of 15. In other words: the point would be plotted at (13,15). Let's do it, using a "o" as the indicator's point.

*

*

*

*

*

*

* o

*

*

*

*

*

^

Note - when plotting this new point this way, you can see it "doesn't make any sense".

Now, do it once more, but this time plot the point half way back in "time"

I'm going to us an "X" for my new value.

*

*

*

*

*

*

X o

*

*

*

*

*

^

You see, you found the so called 'average value of Y', but you "need to find the same average value" for the "time" values of 'X'.

- - - - - -

So, now look what happens when you plot the next three values of first the Simple Moving Average, and then that of ARIMA

*

*

*

*

*

*

X o

X o

X o

X o

*

*

^

*

*

*

Just a final note, this ARIMA is supposed to be 99.5% effective in filtering out the sub-cyclic components, from the other cyclic components that are longer than the ARIMA span you choose as your filter. The only thing its nice to do is pick a filtering length - that when you do the (n + 1)/2 step ... you make the result is an odd number. If you do, it plots up soooo much better. - right down the exact middle of this line.

-Bruce Zerr

Can anyone pass me the spreadsheet to do that....