I know examples where adjacent watersheds respond similarly in same physiographic area, and other instances where they are different response.  I think the difference has something to do with geology substrate differences, either karst lenses or fractured lineaments that can under high water levels, move noticable amounts to groundwater or aquifer.  With the cost of transducers, even if they fail once the n a while, you can get numerous stream gauging water levels per day, same with recording raingauges using weight ot tipping bucket.  

But I probably would value the daily rainfall and 6-hour streamgage data over a model.  And streams often give hints or indicators, such as bankfull and channel type indicators that suggest a way to compare response with no data.  Also, if there are culverts, they often leave a wear line, rust line for a low flow, and if you see this line at 1/3 or higher of the culvert diameter, it is likely undersized.  And for snowmelt systems or coastal plain systems with extended hydrographs, 6-hour stream gauging may be OK.  But where there are intense thinderstorms, monsoons, hurricanes, tropical storms, etc., much data is lost with such infrequent measures, especially for small, flashy hydrologic units.  

You may have heard of Luna Leopold (a famous professor, geomorphologist).  He set a staff gauge in a stream near his house, and during storms, he would use binoculars or a telescope to read the water level changes with time, and had his own backyard raingauge he measured regularly (I,assume it was clear, with graduated lines to also read from his window).  If I was suggesting, the 15 minute intervals that USGS uses is generally fine, but on very small catchments, flashy responses, 5 minute intervals may be helpful.

We often get buy with little information and various models, but flood damage to society continues to be high.  Water data is not a loosing investment in my opinion, especially when dams overtop and fail, or roads fail from inadequate bridges and culverts.  We are so technical, yet perhaps not that convincing that investment in water data can produce dividends, and don't get me started, but we are making mistakes in how we are presenting risk in hydrological design unless 50% failure is acceptable 

Thanks William

Very nice penetrating and accurate kind of reaction. each and evry bit you wrote is correct and I 100% agree with your views and analysis. Stream flow marks are the best and most reliable data for hydrology, even surpassing modern days sensor based or other hi-tech dynamic modes of gauging flow levels either in stream or over some kind of hydraulic control structure (weir/flume etc). For any kind of hydrologic signature from any natural catchment, it is the peak flow rates which matters a most , and this peak stage of flow has direct connectivity with peak stage of flow in a channel at any given point of time....and yes..as you have very rightly said  the  peak flow marks on sides of the channel in different looks ( may be a sediment layer marks, or vegetation adherence lines or any scour or cut line in regular manner along side walls etc) are the best indicator to work out the peak stage of flow and in turn the peak discharge rates..Very nice example 'Luna Leopold' and i have generated eagerness to sail across it some time. I too have many such attempts in real ground situations, where in degraded un-accessible  forest catchments, some simplistic locally fabricated crest gauges were devised and installed putting / reusing some waste pipes and erecting them in centre / bed of channels at appropriate location and then putting some kind of light floating grounded substances like saw dust or charcoal powder at every morning and then placing a wooden stick or scale inside pipe, and visiting that pipe daily in morning by taking out the wooden scale and observing the layer of saw dust or charcoal particles whatever used ..a circular layer circumferential was easily visualized at each visit, and setting its height in accordance to channel bed, gave a good estimation of peak flow stages and in turn the peak discharge rates too. 

But still,the core theme or expectation from learned friends here, is that " if we like to simulate full hydro-graph " under given situation ( rainfall @ 24 hrs time step; discharge @ 6 hr time step) what could be the simplistic and alternative ways and options  , and which options to be used to make use of such situation where neither it is a truly gauged situation nor a truly ungauged situation. Our intention is to predict full hydrograph under varied rainfall input conditions and input rainfall is available only at 24 hours time step. while we wish to conduct calibration and validation for discharge hydrographs having time steps less than 24 hrs ( may be 3 or 6 hrs).  

for tropical situation like India, even when the base of hydrographs is in days ( as in this case), the exact duration of input rainfall (if taken as 24 hours) is full of errors/uncertainties, as exact rain or storms in real situation which are responsible to produce overland flows/channel flows are hardly of few hours (not 24 hrs) and it has great impact on shape, volume and overall hydrograph function. so at last  it could be termed a work on an ungauged catchment only   , even tough we have some data (partially guaged)..  How to harness the potential benefits of partial observations or how to transform time scales to make them identical? ( conversion of 24 hour rain fall data into 6 hour rainfall values; or taking 6 hours discharge date into a 24 hour time step value- both cases have errors/high uncertainties which are not wise to get recommended for predictions in nearby ungauged catchments ..whatever the kind and model we adopt....

Once again thanks for your nice words and giving time on mine question..

Regards

Gaur

Greatings Gaur,

Analysis with 24 hours should be done with great care, because the storms have durations of 2h, 3h, 10h etc. Moving to 24 hours, is equivalent to introducing a random variable D = 24-DT, where DT is the rain duration, which increases the problems in models calibrating. It is worse if the storm duration "DT" is small. Up to 6 hours in Quito-Ecuador it is possible to transfer information from 24 hours to 6 hours or from 6 hours to 24 hours.

I share with you this graph. Is a result of my work, that has been calibrated. Soon I will publish my first book "What the storms hide". " Lo que esconden las Tormentas" in spanish. I hope to publish it this week in Amazon. I hope I've helped.

Best regard.

Thanks a lot Franklin,

It is just a superb answer for me, which I was seeking since long. Wonderful approach you have mentioned. Congratulation for your great book on a great issue in Amazon. Wishing you all the best and to be honest I am happy to get attached with you here anticipating futuristic help & exchange of knowledge / support of you...

thanks again..very nice and most relevant clue you have given....

Regards

Gaur 

I would not consider a gauge where you had 24 hour rainfall data as ungauged. There is a lot of information in that dataset that will help you understand the behaviour of a catchment. An example of an ungauged catchment would be one that doesn't have a streamflow gauge. A more extreme example would be one where you don't have any rainfall data either. The data you have for the catchment should reduce the uncertainty in the predictions since you have information to use, but not necessarily the difficulty in putting a model together. The problem would be if the uncertainty in the data (how well the data reflect the actual rainfall when binned to that temporal resolution) is so large that the information gained is swamped by the uncertainty.

Given the hypothetical situation of a 2000ha catchment, with daily rainfall and 6 hourly streamflow data, you would be missing a lot of information on rainfall intensity, as well as the streamflow. Unfortunately, with the data already sampled, there is no way of regaining that information. You can improve things by adding additional information. For example, over a decade ago, I had a student looking at using a data mining approach to see if it was possible to capture intense events using other climatic data (temperature, wind speed, ...). This won't give you a finer temporal resolution rainfall time series, but at least will help constrain the possible sub-time step rainfall patterns. Another idea is to look at the spatial variability in the rainfall (e.g. standard deviation of the rainfall on a particular day across all gauges). If the spatial variability is high, there is an increased probability of a more intense (i.e. shorter duration) rainfall.

There will also be missing data from the streamflow record from having 6 hourly data. For a catchment of only 2000ha, the peak flow from an intense rainfall event will be significantly undersampled (how much depends on the characteristics of the catchment).

Thanks Barry,

very much correct you are.

Indeed same were the set of questions/doubts in my mind and still exists. And this is not the only situation with me, or at this place, or for hypothetical 2000 ha sized catchment or 6 hr time step. in the real sense it is the prevailing set of conditions in a big number of catchments...so it remained a big problem. we have to initiate its solution treatment at some point...sometime....to harness fruits from such partially gauged conditions if it is feasible...I don't know correct or not..but I am extremely overwhelmed by the efforts/approach initiated  by Franklin, atleastsome thought is needed on this issue...

One thing you have correectly said that for smaller catchments like of 2000 ha or so , 6 hr is even not the correct time step , as the time of concentrtionof such catchment is hardy an hour..so rainfall intensity for duration equal or less than 1 hr will govern the exact magnitude and timimg of peak of resulting hydrograph/s...

Question is segregation of rainfall data...even may be a stochastic one...