Some definitions would be useful to avoid any confusion in the nature of measuring sites:


A traditional station is defined as a site where discharges are determined from a group of instruments based on what is considered a univocal relationship between water level and discharge.

Thus the station is made up of:

The European norm 1100, part 1, deals with station characteristics.
The ideal site rarely exists in nature; the site chosen must nevertheless meet, to the extent possible, certain hydrological criteria.

1.2.1 Fidelity

All discharge to be measured must run through the section. If this is not possible, we must be able to account for whatever is causing the problem (back water works, overflow of embankment, etc.)

Zones which contain aquatic vegetation are to be avoided.

The relationship between water level and discharge, verified by frequent gauging, should be as much as is possible, univocal, that is, each level should correspond to one and one only discharge.

Long term ground stability of gauging sections is one of the most important criteria sought and may possibly require construction of a sill.

En theory, the flow in front of the station should be:

A current is considered permanent when its hydraulic characteristics remain constant in a given section as a function of time. It is considered uniform if, in addition, the velocity vector constant along every streamline.

As a practical matter, rivers donít always meet these conditions.

That makes it necessary to find a control section with a constant shape and where the approaching current speed is slow, followed by sufficient acceleration so that all changes in the flow downstream will not have any effect on the upstream bank of the control section.

An upstream control can be:


Artificial A naturally occurring narrowing is a preferable location: the up stream risks are minimal and it is sure to be permanent.

It is often difficult to find a site with a single control section. A rating curve is then constructed by integrating the changes of the control. When the natural conditions are such that a natural control cannot be found, the solution is to build a sill. (artificial control).

1.2.2 Sensitivity

So that the control station can detect small changes in discharges, slight variations in flow must result in changes on the gauge that are large enough to be read. The "basic unit" can be considered as the flow variation that corresponds to a movement of 1 cm on the scale, a change that can be easily seen.

The sensitivity of a station is further enhanced when the roughness is important, when the slope is gradual, the depth is big enough and cross section is not wide.

The sensitivity can be quantified using a rating curve . The study of variations in discharge show significant differences among low, medium and high flow conditions and, depending on the morphology of the stream bed.

To evaluate the sensitivity of a potential site, making a sketch of the rating curve permits you to anticipate certain other problems; installing a staff gauge and doing a few discharge gaugings will help to identify the constraints of the site.

A quantification trial on a few selected stations in the Seine and Maine basins resulted in quantifying the sensitivity of one station by breaking out the changes in discharge values that corresponded to a 1cm change in water level for a battery of hydrological parameters. The discharge characteristics used were:

The average sensitivity by category is the geometric mean of the sensitivities found for each of the two characteristic discharges of the category.

The proposed limits are provided by "expert" experience and ought to be validated by a more methodological approach.

Site rating not very sensitive sensitive very sensitive

A good sensitivity in low flow conditions translates into noticeable changes in water level when there are slight changes in discharge, which is not easy to obtain. Also note that sensitivity falls with the first overflowing in flood conditions.

1.2.3 Accessibility

The station should be accessible in all circumstances especially during flooding. A few simple site improvements can sometimes make conditions more practical (especially access roads to the station as well as the gauging sites).

The distance between where you park and the station itself, safety conditions at the station and at the measuring sites and road accessibility are factors which have a considerable effect the daily management of such an operation.

An ideal site with difficult or dangerous access is not as good a choice as a site that, although a little less satisfactory in hydrological terms, will be used on a continuing basis.

1.2.4 Technical constraints of a site

The station should be designed to remain operational even in extreme flooding. An inquiry about the highest known water conditions will allow you to locate the equipment out of flooding range.

The staff gauge should be located as close as possible to the sensor (e.g. in the case of wells, the oil sump filter acts as the sensor) and installed so that it is easy to read at all levels, is readily accessible for maintenance and will not be disrupted by floating debris. The zero level of the staff gauge should be hooked up to the NGF system and to one or more well identified local markers.

The sensor should be located so as to avoid possible backwash and buffeting; A sunny spot is often favorable for the growth aquatic vegetation.

Stagnant water zones often lead to problems of silting up and create a need for added maintenance and sometimes preclude the use of certain techniques. Electric and telephone services are a definite asset. It is a good idea to ensure protection against electromagnetic interference. A study and on-site test measurements are recommended before the installation of equipment for a modern station, (for example, which automatically transmits its data). Finally, there could be technical constraints specifically associated with each sensor that would lead to the choice of equipment based on the site as well as special civil engineering work adapted to the site.



It is sometimes impossible to find a stream section where water level and discharge are directly linked, notably when the staff gauge is located in the backwash of a mobile dam. The same reading on the gauge could be associated with different discharges.

Calculation methods used for determining the discharge are derived from Stricklerís formula:

V = K x Rh2/3 x I 1/2

Where V is the speed in m/s, K is Stricklerís coefficient, Rh hydraulic radius in m and I is the slope.

The discharge is determined by measuring water level at two points which allows us to define the slope. Once the cross section is recorded, the hydraulic radius is subtracted from the level measurement; Boyerís method is basic method used in HYDRO. Other formulas derived from this method are used by some station managers.


2.2.1 Choice and location

The 2 stations should far enough apart to allow a satisfactory measurement of the difference in water level between the two gauges. This means the choice of a sufficiently long reach in the case where there is slight slope. The stations should be located a few hundred meters from either end of the reach.

The flow in the reach should not be subject to major disturbances. For this it must have regular geometric proportions all along its length. To be altogether avoided are sections with pools which fill and empty in an irregular fashion or tributaries which flow into the reach. Also avoid areas where dominant winds blow along the same axis as that of the reach.

The upstream station usually has the reference gauge and also serves as the current meter gauging section. This means easy access to this measuring section, flow regularity in the immediate vicinity of the station itself, a regular crosscut profile and straight, clean banks.

2.2.2 Set up for measuring slope

Staff gauges

The upstream and downstream stations should have a staff gauge that is clearly readable and protected from the wind (the effects of buffeting)


The correct installation of sensors depends on the precision of information, therefore the calculation of the difference in water level between two points of measure.

No matter what, the sensors should inside or under vertical stilling wells connected to the stream and in close proximity to the staff gauges and high enough to be protected from high flow. The wells must have a minimum diameter of 1m.

Regardless of the style of sensor, it must be accurate to 0.5 cm. It is advisable to monitor the synchronization of clocks and proper simultaneous operation of mechanisms.

2.2.3 Operating precautions

A twin gauge station is generally set up on a river with a slight slope. Staff gauge readings demand close attention and should be taken carefully, avoiding parallax errors.

If you use a paper limnigraph, (1:5 scale), here are a few precautions:

To permit the correct setting of the sensors, each station should be equipped with a sounding support overhanging the wells; its altitude will be determined relative to the N.G.F. Thanks to this system, the level of the water in the stilling well can be measured to the millimeter and the slope can be very precisely calculated. To facilitate this operation it is a good idea to cut off the flow from the river to the well using a valve on the water diversion tube.

For the gauge setting operation, be careful that the flow in the reach is not disturbed at the moment of the measurement. Particularly in low flow conditions, in the case of a navigable river, it is essential to take these measurements during non-navigation hours to avoid the undesired effects of boat wakes.

These same recommendations are valuable for flow measurements to check the current meter characteristics. They should be done during non-navigation hours and when the water surface is calm (generally very early morning).


The precision of discharge measurements depends essentially on the precision of the water level readings and setting the sensors.

However, the lower the difference in levels between gauges, i.e., in periods of weak flow, the less precision you will have. In other words, the uncertainty of the value given becomes more uncertain.

The multiple variations affecting a body of water do not allow the determination of instantaneous discharges; the default time scale is one day.

In a period of low flow, recording the wind speed gives an added dimension to the quality of the data.


The method of calculation comes from Stricklerís Formula, assuming that the coefficient K is constant. If the river is not dammed in its low-flow channel, the flooding field will make the value of K change. Therefore, stream gaugings are still imperative.


Technical evolution is moving towards the disappearance of twin gauge stations, increasingly replaced by ultrasound velocity measuring stations, the main obstacle to this change being economic. This evolution will also provide more accurate discharge measurements in low flow conditions.


Stations which directly measure stream current speed by ultrasound waves were developed and put into service in Switzerland in the early 70ís. They were then used in other European countries (UK, Germany, the Netherlands). It wasnít until the late 70ís that the first such stations were installed in France: Paris Alexander III on the Seine and the "Canal du Nord". Currently, some 15 stations are operating on French rivers: CNR, DIREN Ile de France, Nord pas de Calais, EDF DTG, SN Strasbourg and Lormines. Stations of this type are widespread in the area of water treatment (see DDE 93, GEMCEA)

The main manufacturers are: Stork (the Netherlands), Atlas Krupp and Ott (Germany), Ultraflux and CR2M (France), Peak-Martac (USA), AFFREA-Hydrosonic (Canada).

This kind of measurement is interesting because it permits us to calculate the discharge even if the water level-discharge relationship is not unique (afflux due to an impoundment or a reach subject to tides). Ultrasonic measurements provide the parameters needed to calculate discharge, i.e., the water level and the average current speed at the depth of the measuring cord. And, the values displayed by the ultrasound sensors being algebraic, this system can also determine the direction of the flow.

The major interest of this type of station is, therefore, the ability to measure discharge where classic gauging methods cannot. But contrary to a widely held idea, this mode of measurement still needs to be validated in very low speed water conditions. Indeed, this type of station provides only average current speeds over one or several horizontal lines of measure. Traditional gaugings are still necessary to establish a relationship yielding discharge as a function of the water level and the speeds measured, so it is imperative that it be technically feasible to perform gauagings over the whole range calibrations.


Measuring water speed by ultrasound is based on measuring the time it takes impulses to travel back and forth between paired transducers and water in movement. The difference between impulse sent and impulse received can indicate the direction and the speed of the current. The line connecting the two transducers is inclined at about 45° form the presumed direction of the flow.

The sensors can be configured to handle current flows that are not parallel to the riverbanks.


You must find a site that is easily accessible to reduce the heavy cost of landscaping and facilitate station maintenance. It is also a good idea to choose a section where the streambed is stable in order to avoid frequent corrections of the parameters used to calculate discharges.

In the case of a single measuring line system, be sure to choose a site which does not have bends or direction changes, otherwise the measurements of average speed will be false.

The presence of obstacles in the current can affect the sound waves emitted and thus alter the measurement. Besides the ensuring regularity of this section, you must pay attention to aquatic vegetation in the waterway and watch for large amounts of floating debris. In addition, donít forget to take into account the effects of river navigation; when there is a lot of traffic, dispersing air bubbles can greatly perturb the measurement.

Ultrasonic measurements are inadequate for rivers where the concentration of suspended solids is greater than 1.5 g/l during some flood conditions because there is too much signal attenuation.


Five years of experience operating EDF DTG ultrasound velocity measuring stations have yielded the following principles of installation:



The Starflow functions as an emitter and at the same time as a receiver of ultrasound frequencies. The current speed is measured thanks to fine particles suspended in the water and which act like moving mirrors. The sound wave undergoes a variation in frequency because of the movement of the reflecting particles. This variation in frequency allows the speed and direction of the current to be calculated. Likewise, reflections of sound waves on the surface of the water allow the computation of water level.

Knowing the cross cut profile of the section, it is then possible to calculate the wetted section and, multiplying it by average speed, figure out the discharge.

The only station currently using this type of set up is DIREN Basse Normandy.


Place the central acquisition unit on the bottom of the stream in a regular section and, if possible, in an artery with regular flow characteristics. It is best to attach it to a mobile platform which is lowered into the river along a line attached to a stake planted at least 1/3 of the way across the width of the section, though this could get in the way of river traffic. The central unit is entirely submerged; only the battery and the modem remain out of water on the bank. Any adjustments to the Starflow require it to be brought out of the water. The variable length power cable and RS 232 cable which link the central unit to the battery and modem are fixed to a chain that is laid on the bottom. Installation and maintenance are rather complex and time consuming operations but this device yields reliable and coherent results. Current speed and discharge data, should be initially validated , of course, by gauging.


This type of equipment allows discharge, water level, direction of current in estuary zones to be measured with good sensitivity. It opens up, therefore, interesting possibilities to resolve some situations which current can not be addressed with other traditional techniques.

The equipment tested in 1995 at DIREN Basse Normandy has been totally satisfactory. The two stations purchased in 1996 had some communication problems, notably for retrieving data files. These kinds of problems can probably be fixed quickly because it has to do more with software than hardware.


A detailed topographical survey of a prospective site is highly recommended. This should be comprised of three crosscut profiles: upstream from, directly at and downstream from the measuring section. If necessary, it should be completed by a stream profile. These surveys will allow you to determine the slope of streambed, the cross section, control and hydraulic characteristics, overflow points of each bank, etc.

This survey information should be organized and added to the site file. Its use in determining the rating curve is essential. In extrapolations, the absence of topographical information can lead to serious errors.

Likewise, photographs of the station site and the measuring section in different water conditions will allow you to envision the various flow circumstances possible.


Hydrometry depends on written information in as much detail as possible, so as to mitigate our imperfect memory (the 5th gauging of the day could be confused with the 4th when we are assembling data two days later) or changes by the recorder or other persons involved in the process.

Coming back to old data is also one of the characteristics of hydrometry. In this case it is necessary to have data that is as complete as possible on the measurements concerned, to ensure reliable interpretations resulting in quality discharge information.

In this sense, we in hydrometry have long been exercising quality assurance measures.

The same practices apply to hydrometric stations. The totality of data gathered is combined into a "station file" which is more and more complete to the extent that each staff member contributes additional information.

This file should include:

The updating of this file is often considered a tiresome chore by station personnel. It is invariably of great usefulness whenever old anomalies are found which require data to be reanalyzed. For the hydrologist, the difference between trying to conduct a historical study with uncertain data and in which the witnesses have often disappeared, and having at hand the necessary data, clearly annotated, is considerable.



When the recorded water level measurements at a station are doubtful or haven't been made in a long while, the station personnel should take actions to fill this information gap, created in the continuous chronicle of water levels as a function of time (S/T).

The action necessary to rectify this absence of information is called "reconstitution".

The methods and difficulties involved in reconstitution are quite varied, depending on the amount of time for which data is missing. It is possible, nevertheless, to give a general idea.


The easiest cases are usually on rivers with no external influences, concern relatively short periods (at most, 3 - 8 days), and occur during a time without rain or flood conditions.

In this kind of a case, it is often sufficient to retrace the water level variation curve as a function of time, starting from existing data, then transform the estimated water levels into discharges. This kind of situation should be considered an exception; reconstitutions should generally have to do only with discharges. The existence of a "doubtful water levels" code in HYDRO facilitates the transparency of this kind of reconstitution.

Careful! With the increasing incidence of "water uses" (irrigation, various draw-off, etc.), reconstitutions in low water conditions have become more delicate.


Reconstitutions are the most difficult when there is influence, especially contributions linked to rainfall. Here, the means needed for reconstitution are more consequential. Likewise, a long period with no information, several weeks, for example, requires a deft touch. If it has rained significantly, the reconstitution is all the more complex when the stream current reacts to the precipitation and is not as regularly fed by groundwater systems.


The first rule in this type of case is to find, other than discharges of close-by measuring stations, available information on the rain that has fallen (the levels observed in the proximity, etc) that serve as elements for the basis of reconstitution.

It is essential to keep a record of the elements and methods used for reconstitution of missing information, even if the means used aren't perfectly orthodox.

With certain paper recording instruments, when the paper advances momentarily at a speed far less than normal, it is possible, using the "compressed" graph to reconstruct a plausible graph line, paying close attention to the peaks and gradients corresponding to rising and falling water levels.

In the case of a clock breakdown or forgetting to reset the stage recorder (other limnigraphs are also concerned), knowledge of the extreme values previously recorded is obviously invaluable since that gives you a range of water level variation during the breakdown, under the condition that the stylus remained in its normal position.

If there are no indications about the range of water levels attained during the absence of information ( failure of electronic recorder e.g.) the rule becomes to use data from one or more comparable measuring stations, taking into account the amount of rain that has fallen in the basin concerned. The data teletransmission becomes interesting - it could supply measured data at regular step as well as the information concerning the equipment operating.

The following are considered comparable in terms of hydrometric stations:

If the time period of missing information is not too

irregular or too long, it is then possible to envisage a reconstitution based on a correlation between discharges taken from the station in question and those of a "reference" station or stations.

Note: It is usual to see the comment, "discharges reconstituted by graphic correlation" on daily discharge tables. This type of reconstitution is probably satisfactory for cases such as those described previously in the "easy" category. Still, it would be preferable to use regression analysis to correlate the existing streamflow parameters related to the station in question and the reference station, if the reconstitution covers a significant amount of time.


The variable to study can be:

where n = 5, 10, 30

It is preferable to use variables belonging to the same hydrological year and during the year of reconstitution. The use of variables from other years can be used to complete the approach. Beware of the existence of differences in discharges of "similar" rivers that could be the result of different climatological histories (depending on the year) of the respective basins.

For a correlation to significant, the range of discharges considered must be sufficiently distributed and framed within the range of those found at the reference station during the period concerned. The possible seasonality of the correlation will also be verified.

The formula obtained must make physical sense:

If: Q unknown = a Q known + b

a must be coherent with the comparison to drainage basins

b must be low compared to the range of discharges to reconstitute. Here again the geology could influence the principle. A basin flowing into permeable formations added to a basin flowing into impermeable formations could justify a large value for b.

When the area of respective drainage basins is too different, it often becomes necessary to use a logarithmic correlation of discharges.

A correlation formula that does not make sense must be rejected even if the coefficient of correlation is very high. Beware of samples that are too limited or too asymmetric which could lead to incomplete conclusions.

Visualisation of the points distribution of correlation is indispensable. For low and medium discharges the 'zoom' tool is to be used.

It is advisable to examine the evolution of the representative points of correlation before and after the period of reconstitution. The significant differences between points, if they exist, could be due to variations in rainfall in the two basins or a difference in reactions to the rainfall in the basins that the formula doesn't reflect. In any case, the application should be used with care.

A visual examination of the reconstituted hydrograph will be done. If necessary, it can be manually optimised to comply certain time delays and to maintain coherence in terms of the total volume of the streamflow.


If it is not possible to reconstitute them, it could be interesting to try to evaluate peak discharges of floods during the period of missing information in order to have a continuous annual series of maximum discharges (for calendar year or hydrological year). This becomes more delicate as drainage basins become smaller.

First you should look in the field if there are high water marks on permanent objects. You must be careful about marks left on trees or branches whose position might have been changed by the speed of the current.

In the absence of any such marks, there could be a possible correlation between known peak discharges at the "reference" station and those of the station in question; be sure not to forget to examine rainfall figures for the two drainage basins also.

When these stations are not located on the same river, or if their respective drainage basins differ greatly in surface area even if on the same river, there is a good chance that there will be a very loose correlation or that the estimation of the desired peak discharge will be in quite a broad range. Estimations derived this way can still be of value in situating unrecorded high water discharges in a sample of maximum flood conditions (annual maximums or flood levels above a certain threshold).


In certain cases, especially concerning extended basins for which there are hydrological models, we could see the use of rainfall - discharge relationships to do reconstitutions (the less permeable the basin, the easier this would be). Or, you could correlate discharge data between the station with missing data and other upstream stations, taking into account, if necessary, the propagation effects between stations.

The use of daily hydrological models (Cemagref GR4, EDF's MORDOR, for example) allows reconstitutions from rainfall. The difficulty comes in setting up the model, which requires considerable time as well as the integration of rainfall data itself.

If the absence of information is long and includes a particularly rainy time, you can assume it is not possible to make a very detailed reconstitution. Only the monthly discharge figures will be estimated. In this case, if seasonal variation is observed in the correlation, then analysis of correlations will be limited to the monthly discharge for one period of the year instead for the entire year.

An example:

In Ćuf, in the upper parts of the Essone near Bondaroy is a groundwater outlet from an aquifer with considerable inertia. The discharge from this station was considered to be the sum of base flow (relatively stable over time) plus runoff from rainfall in the Pithiviers area.

The reconstitution is therefore done by extending the base flow recession curve and, when necessary, adding in the rainfall data calculated according to the actual rainfall.


One imperative rule concerning reconstitution is to express your values in only two significant digits, generally speaking (4.7 instead of 4.68). You can make an exception when dealing with discharge values between 1 - 2 m/s; in that case the last digit should be 0 or 5 (1,3 instead of 1.32 or 1.95 instead of 1.94).

A reconstituted water level or discharge should always be identified as such by using a validity code of 8 in the HYDRO databank. If the reconstitution is uncertain, don't hesitate to label it with a code 5, "doubtful".


The chain of data acquisition that includes a sensor, a code converter and a recorder is not totally safe from errors in interpretation. This system generates information that must be analysed, critiqued and corrected. These different steps don't occur in chronological order but must be repeated as necessary throughout the process, from the site survey up until entering the data into the hydrological databank.

The ultimate result, the discharge value, is, in turn subjected to a validation process. The necessary rigor of each step allows you not only to select most significant data but also to guarantee the integrity of the final result.


Setting parameters for the acquisition chain in terms of sample frequency, precision and intervals between memory updates is something that must be adapted to each river. It is useful to take measurements at different times according to water levels, as a function of each site's own characteristics.

The file of accumulated data contains a considerable amount of information for calculating the discharge, which makes it necessary for the subsequent detailed processing of that data.

The first check is on the terrain. The congruity between times and water levels for both the measurement sites and the water level recorder are compared. If necessary, essential adjustments are made. Then, verification of the parameters is done. A check of the linmigraph on record is made to detect any possible anomalies. If there are any, they are noted on the station log with a detailed account of actions taken to resolve the problem (battery changed, reset a recorder or gauge, etc.). This log contains the date, time, staff gauge level, all essential parameters as well as any notable changes to the station in a hydraulic context. Then, in the office, you will proceed with the normal processing of data.

Whether you use the graphic recording or digital recording, it is imperative to save the raw data, without any time limit, on a reliable medium. CD-ROM recorders now provide excellent low cost long-term storage solutions.


After identifying any recording anomalies that have been identified, whether caused by a discreet event or progressive equipment malfunction, corrections must be made. The basic verification is done by lining up the limnigraphs' beginning and end water levels and comparing the recorded values in between.

The "cherrying" is an intermediate limnograph processing step. The cherrying unit of measure is the mm. The curve is coded in a way so that the recording between two "cherries" can be reasonably assimilated into a straight line. The distance between a straight line defined by two successive cherry points on the curve and the actual curve should never be more than 5 mm nor should it generate a discharge error margin greater than 5%. In the case of significant variation in water levels, increase the number of cherries.

Graphic reversals on water level graphs should be analysed with precision and correctly entered/identified. The beginning and end water level values are indispensable. If there are multiple graphic reversals, the limnograph should be digitalised and carefully examined, which should allow you to find the error since the recording characteristics of raising water levels and falling water levels are different.

It is a good idea to monitor the continuity from one recording sheet to the next to avoid "stair stepping" of data limits.

If this step seems to simplify the curve, it is advisable nevertheless to pay close attention to even slight variations, which must be taken into account.


With the reduction in cost of data storage, it is much easier today to keep even moderate variations in amplitude. Good knowledge of the site is essential to judicious processing of the data. The following different steps can be differentiated in the processing.

Filtering consists of eliminating background noise linked to the accuracy of the measurement of the waterway as it is effected by surface turbulence, or slight, random, insignificant variations of a few seconds. The important thing is that the department is aware of the technical rules used by each site and that they are written and applied in a consistent manner. The amplitude of background noise is modest ( up to a few mm).

Next you proceed to the smoothing out of the slight variations in water level which account for only a modest volume, which are more or less in equilibrium and which don't last more than 30 min on average. The source frequency is thus longer and the physical cause can be identified (example: a 'constant' levelling valve). Smoothing should be the result of a careful thought.

Examining the limnograph can reveal isolated or random values that should be eliminated. You should also think about reconstituting certain values in order to have a continuity of discharge (short time gaps in data or delay in water level, etc.).

Compacting is another step that involves breaking up the curve into discreet straight-line segments. This makes a far greater number of pivot points than if the process is done manually. This is the equivalent of the 'cherrying' process. The distance between lines defined by two successive cherries and the actual curve should not be greater than 5 mm nor should it generate a discharge error of more than 5%.

The visual inspection of the digital recording paper is extremely useful, as are the raw hydrograph record and the processed data which could validate or invalidate some simplifications.

All of today's information technology processes should facilitate the operator's job and should complement his expertise but should not take the place of physical examination of evidence.

Some software programs propose a smoothing function, automatically transforming raw data into average values by using a filtering algorithm for a wide range of discharge values. The absence of evaluation and control leads to masking significant information such as the beginning of a flood or the length of a maximum value. We advise you not to use this type of procedure.

Whether for graphic or digital files, you should always verify that you are working on the right station; it's always possible to switch data sheets by mistake.

Raw water level is what we call the limnograph recorded at the site.

Validated water level is what we call the processed results described above.



All during the processing, evaluation has played a part in correcting recording anomalies in the field acquisition chain.

A certain number of verisimilitude checks are should be done to detect divergent values by identifying outliers. Thus, a value of 10 m will be abnormal when the range of fluctuation is between 0 -4 m.

Comparing a gauge reading and the corresponding limnigraph will allow you to identify anomalies. Recording errors are possible and reading the station log can yield a lot of information if it indicates work being done on the river, obstacles or any other significant changes to the station.


The validated water level value entered into the databank is always the same as the raw recorded water level value.

This principle gives rise to some very specific adjustments. On a watercourse that is seasonally influenced (by vegetation growth for example) and given a sufficient number of gaugings, it is possible to put a correction of variable water levels into the databank, allowing us to keep a validated water level and rating curve.

If an obstruction temporarily disturbs the recording of a water level, with a calculable influence on it, it could be permitted to enter a corrected validated water level value as long as it is accompanied by code "8". If there is a long-term effect, then the rating curve will need to be changed.

Let's consider the case where a reach becomes completely dry or where natural discharge is interrupted for only a few days per year and then only once every n years: a brief increase of discharge followed by a drop in discharge during refilling. It is then necessary to create two stations: one with the observed discharges (and the validated water levels) and the other with reconstituted discharges, to use for statistical calculations where the discharge data has been smoothed and spread out over the period. This particularity should be signalled by comments for the two stations.


There should be an evaluation of the discharge values themselves.

A visual examination of the discharge hyrdrograph is a very useful exercise as it is easier for the hydrologist to detect an anomaly in a graphic than in a table of numbers.

This examination should be done at different time intervals- monthly, quarterly, annually- and special attention should be paid to marks. The transition from one limnigraph to another, from one rating curve to another, from one calendar year to the next are all classic traps. The level of precision should obviously be adapted to the time scale used, instantaneous or daily.

Scale changes are sometimes necessary: a groundwater recession anomaly is difficult to detect on a monthly scale; a coding error is easier to notice over a short time period.

The alternate use of arithmetic and logarithmic scales will permit you to see many kinds of anomalies. The graphic possibilities that we have at our disposal today give us a golden opportunity to look at our old data where we will surely find a number of discreet but non-negligible errors.

In terms of managing low water conditions, a new problem is a rapid evaluation of data. To aid the hydrologist, we advise the calculation of recession curves from the last reliable data to validate the new data in real time. In areas that have been affected by draw-off, each case of water draw-off should be identified by type (retaken of irrigation after a rainy period, for example) which will complete the classic approach on natural recession.


Comparing data between stations is one good way to identify anomalies. Comparison does not mean similarity; it's in terms of major and cumulative differences that this analysis should be made.

The superimposition of hydrographs allows you to make simultaneous visual evaluation. The HYDRO databank authorises intensive use of this faculty with the possibility of a time lag scale.

simple test often used to judge if there is a problem with a station is called double mass curve. You graph the cumulative values of daily discharges for station Y that you want to evaluate against the same values for reference station X. A break or discrepancy in the one of the curves is a sign that one of the stations underwent a sudden or progressive change at a certain date. (This technique can also be used with flood water levels.)

It preferable to use the residual mass curve method which consists of two series X and Y linked by a single correlation Y = aX + b which defines the curve of cumulative residuals.

If the curve Ri exceeds the limits of the confidence interval, there can be discontinuity and i provides us its presumed date. This method is available in the HYDRO databank under the heading CUMUL.

Another universal test consists of drawing the distribution of stations in a given area identified by the following co-ordinates:

The examination of the distribution of points allows us to identify the anomalies. The same graph with QIX and log S is equally instructive.

This test is also available in the HYDRO databank under the heading "ZONAGE".

The use of stations located at a confluence of two streams can also allow you to identify anomalies: if the sum of the two discharges is greater than the total discharge downstream, a close analysis of the three discharges is in order, either their respective rating curves or their limnographs.

Likewise, verification of time lags upstream and downstream can cause problems if the inversion is constant or systematic for a given frequency.

If the setting up a permanent evaluation system for the station seems excessive, self examination of the station is nonetheless imperative.

Anomalies observed in the area of discharge values go back to initial data: water levels, gaugings and rating curves.


It is after the evaluation that the validation is done.

If production of good preliminary data is possible in a short time, you should still recognise that the validation of the data must be done with perspective and without haste.

Certain verification techniques mentioned earlier need ten years of data to be implemented properly.

It's not that this step should be put off indefinitely but a definitive value is never the current one.

This is why it is necessary to continue with overall revisions of historical data, taking advantage of the latest computer and graphic technology. Erroneous historical values in a long series will, if not adjusted, undermine statistical hydrology variables. A missing piece of data from extreme situation can have an even more dangerous effect on the eventual statistical validation of a phenomenon.

The monograph

The monograph consists of a written analysis of the past 10-15 years operation of a measuring station, including all the measurement data, gaugings, hydraulic characteristics of the site, etc. This is done methodically and calmly without imposed time constraints and it should always result in: