The reliability of suspended sediment load data

Transcription

1 Eroion and Sediment Tranport Meaurement (Proceeding of the Florence Sympoium, June 1981). IAHS Publ. no, 133. The reliability of upended ediment load data INTRODUCTION D. E. WALLING & B, W. WEBB Department of Geography, Univerity of Exeter, Exeter EX4 4RJ, UK ABSTRACT The growing ignificance of upended ediment load data ha directed attention to the reliability of meaurement procedure and publihed data. The efficiency of ampling equipment and laboratory technique ha been tudied by many worker, but much le attention ha been given to the problem of aeing long term load. In the abence of detailed record of upended ediment concentration, a number of indirect method of etimating uch load baed on interpolation and extrapolation procedure have been employed. Thee may lead to eriou under- or overetimation of the actual load. A detailed aement of both the accuracy and preciion of load etimate produced for the River Creedy in Devon, UK, uing thee procedure i preented. La validité de donnée relative aux charge olide en upenion RESUME L'importance croiante de donnée relative aux édiment en upenion a attiré l'attention ur la validité de technique pour la meure. L'efficacité de l'équipement pour la prie d'échantillon et de technique de laboratoire ont été étudié par pluieur chercheur mai moin d'attention a été apportée aux problème concernant l'évaluation à long terme de débit olide. Faute de donnée détaillée ur le concentration de édiment en upenion, pluieur méthode indirecte, fondée ur le technique d'interpolation et d'extrapolation, ont été employée. Ce méthode peuvent ouetimer ou uretimer érieuement le donnée véritable. Une évaluation détaillée de l'exactitude et la préciion de évaluation effectuée pour la rivière Creedy en Devon, RU, en utiliant ce technique et préentée dan cette communication. Scientific invetigation of upended ediment tranport in river have now been undertaken for more than loo year. Today, data are available from well over 1500 meauring tation cattered throughout the world (cf. Walling & Kleo, 1979) and the expanion of meaurement activitie ha been paralleled by an increaing need for information on upended ediment load. The growing ignificance of upended ediment load data, e.g. in the evaluation of nonpoint pollution, ha inevitably 177

2 178 D.E. Walling & B.W.Webb directed attention to the accuracy of aociated meaurement technique and to the reliability of available information. Studie of the efficiency and improvement of ampler deign are well known and the US Federal Inter-Agency Sedimentation Project (e.g. FIASP, 1963) mut be recognized a having played a pioneering role in uch work. The efficiency of laboratory technique employed to determine the concentration of individual upended ediment ample ha imilarly received attention (e.g. Dougla, 1971) and national and international tandard relating to ampling apparatu and technique and to laboratory procedure have now been etablihed. To many it might, therefore, appear that the major problem aociated with obtaining accurate meaurement of upended ediment load have been overcome, and that publihed data poe few potential problem in term of reliability. The problem Such a view i perhap jutified in relation to the meaurement of intantaneou upended ediment load in a cro ection, although the etimation of upended ediment tranport in the "unmeaured zone" cloe to the river bed, and the need for a clear ditinction between load relating to total upended olid and thoe relating olely to the inorganic fraction introduce ome uncertaintie. However, many problem urround the aement of longer term load and annual upended ediment yield. Thee problem centre around the marked and rapid fluctuation in upended ediment concentration exhibited by many river. The accurate aement of upended ediment load for pecific period neceitate detailed record of ediment concentration which may be combined with the record of water dicharge. Continuou record of tream dicharge are readily available at mot meauring tation, but an equivalent record of ediment concentration may be difficult to obtain by a programme of manual ampling. On large river it may be poible to collect ufficient manual ample to define meaningfully the record of ediment concentration during period of fluctuating concentration, and procedure have been developed to ait in etablihing the trend between individual ample (e.g. Porterfield, 1972). On maller tream, concentration fluctuate more rapidly during flood event and operational contraint may limit the frequency with which manual ampling may be undertaken. Coniderable uncertainty concerning the accuracy of ediment load calculation may reult in the latter ituation and alo on larger river where ampling i relatively infrequent. Attempt have been made to reolve thee difficultie through the development of equipment and intrument capable of automatic collection of ediment concentration data. Automatic pumpampling equipment (e.g. Walling & Teed, 1971) ha been ued in many tudie and continuou recording turbidity meter (e.g. Fleming, 1969) and nuclear probe (e.g. Rakoczi, 1976) have alo been uccefully employed. Difficultie may arie in relating the point value of concentration obtained with uch equipment to the mean value for the cro ection, but thi limitation i frequently of retricted ignificance when compared to the

3 The reliability of upended ediment load data 179 poitive improvement in temporal reolution obtained. Where frequent manual ample or additional information from automatic ampler or recording equipment are unavailable, the detailed record of ediment concentration cannot be defined and indirect load calculation procedure involving either interpolation or extrapolation of the available concentration data mut be ued. In thi context, interpolation procedure eentially involve the aumption that the value of concentration or ediment dicharge obtained from intantaneou ample are repreentative of a much longer period of time (e.g. day or week), wherea rating curve technique may be viewed a the claic example of an extrapolation procedure. In the latter cae, a limited number of ediment concentration meaurement are extrapolated over the period of interet by developing a relationhip between concentration or ediment dicharge and tream dicharge, and by applying thi relationhip to the treamflow record (e.g. Campbell & Bauder, 1940; Walling, 1977a). The treamflow record may be in the form of either a flow duration curve or a continuou erie, and in many tudie rating relationhip developed from a hort period of meaurement have been applied to treamflow record covering a much longer length of time. Etimate of upended ediment load produced uing thee indirect load calculation procedure may involve coniderable error and the reultant data mut be treated with caution. A triking example of the potential problem i provided by recent work on the upended ediment load of New Zealand river by Griffith (1979) and Adam (1980). Both worker have ued the ame baic dicharge and concentration data collected by the New Zealand Minitry of Work to etimate the mean annual upended ediment load of the Cleddau River, which drain a bain of 155 km 2 in the outhwet of South Iland. Both ued rating curve procedure, but their publihed load of t km -2 year -1 and 275 t km year are different by nearly two order of magnitude. Load preented by the two author for other river in South Iland New Zealand exhibit le marked difference, but thoe cited by Adam average 70% higher than thoe evaluated by Griffith. Similarly, Ongley et al. (1977) have tudied the upended ediment load of five bain in outhwetern Ontario, Canada, and have indicated that load reported by the Sediment Survey of Canada were up to five time greater than thoe obtained uing an alternative data bae and load calculation procedure. The literature contain many other example of major dicrepancie in load preented by different worker for the ame river and which may be accounted for in term of ampling trategy, data availability and load calculation procedure. Equally, a coniderable degree of uncertainty mut urround the likely accuracy and reliability of all ediment yield data produced uing thee interpolation and extrapolation procedure. Although concern for the accuracy of upended ediment load data ha now been expreed by a number of worker (e.g. Dickinon et al., 1975; Loughran, 1976; Walling, 1977b; Olive et al., 1980), it i uggeted that thi quetion require greater attention, both in term of the potential limitation of exiting

4 180 D.E. Walling & B.W.Webb data and the deign of effective trategie for documenting upended ediment load in future tudie. Particular concern mut urround the potential limitation of ediment load calculated uing data aembled for general water quality monitoring purpoe with little or no regard for the pecific requirement of ediment load meaurement. THE RIVER GREEDY CASE STUDY The exitence of a continuou record of upended ediment concentration from the River Creedy in Devon, UK, for a 7 year period, further highlight the nature and magnitude of the reliability problem. Figure 1 preent a comparion of the actual load for the 7 year period with nearly loo load etimate for the ame period obtained uing typical manual ampling trategie and a election of indirect load calculation procedure involving both interpolation and extrapolation. The etimate pan a wide range, and underetimation by a much a 60% i common. The availability of thi record ha further prompted the enuing review of the error aociated with particular procedure and ampling trategie. The data bae Continuou record of turbidity have been obtained from the River Creedy at Cowley near Exeter ince October 1972, uing a light tranmiion enor mounted directly in the river. Although ome worker have experienced coniderable difficulty in etablihing relationhip between turbidity and upended ERROR <%) (a) Interpolation procedure Load etimate Actuai toad inil 1:1 i I ll t y 1 Ml I i 11 [ (b) Extrapolation procedure Actuai load Load etimate uiiiiiii I illii i iiiiniiiiiiiiiiiiii in i \ I^IIILI Etimated upended ediment load (tonne) Fig. 1 A comparion of upended ediment load etimate for the River Creedy obtained uing interpolation and extrapolation procedure with the actual load for the period

5 The reliability of upended ediment load data 181 ediment concentration, a well defined relationhip ha been developed for thi ite. The continuou chart record of turbidity ha been digitized at hourly interval, and the reultant value have been converted to concentration. Value of annual ediment yield have been obtained by integrating the concentration record with the equivalent hourly flow erie. Thee annual load are viewed a an accurate baeline againt which to ae the etimate produced by variou manual ampling trategie and indirect load calculation procedure. The availability of the continuou erie of hourly concentration data readily permit the ynthei of record repreenting different ampling trategie and allow the replication of a particular trategy by uing different ampling time. For example, a weekly interval ampling programme can be replicated by ampling at different time of the day and on different day of the week. Inpection of Fig. 2 provide an indication of the general problem urrounding the accurate aement of upended ediment yield from thi bain uing a programme of manual ampling. Figure 2(a) indicate that concentration in exce of loo mg 1 x only occur for 5% of the time and higher concentration above 1000 mg l -1 are found le than 0.05% of the time. Likewie, the plot of cumulative percentage load v. cumulative percentage of time, derived from a ranked erie of hourly load (Fig. 2(b)), demontrate that 50 and 80% of the load are repectively carried in 0.75 and 3% of the time. Thi ituation introduce eriou problem into any attempt to cover *he major period of ediment tranport by a programme of ampling at regular interval. Similarly, the relatively even ditribution of major ediment tranporting event through the year, evident in Fig. 2(c), poe coniderable problem for etablihing event-baed ampling, when compared to a ituation where thee event are limited to one major period of the year (e.g. nowmelt). The limitation of extrapolation procedure uch a rating curve are alo clearly hown in Fig. 2(d) which illutrate the variation in ediment concentration during a equence of torm event and how that there i no well-defined relationhip between ediment concentration and dicharge. The characteritic of the ediment record hown in Fig. 2 provide a ready explanation of the lack of reliability in the load etimate evident in Fig. 1. In reviewing the reliability of variou load etimation procedure, attention mut be given to both the accuracy and the preciion of etimate produced by particular combination of calculation procedure and ampling trategy. Conideration of preciion i important ince it reflect the conitency with which error may be apportioned to individual procedure and therefore the potential for applying correction factor. For example, Ongley et al. (1977) have uggeted that certain load calculation procedure may underetimate the load of river in their tudy area in Ontario, Canada, but that the rank order of the load will be etimated correctly. The validity of the latter uggetion will clearly depend heavily upon the preciion of the calculation procedure. In the preent tudy, preciion

6 182 D.E. Walling & B.W.Webb -] Maximum recorded concentration (a) Percentage of time concentration equalled or exceeded 100 I: o2 70 =160 H J = S 30 20H 10 o 0 Cumulative percentage of time (b) (C) oo A Fig January 1974 February 1974 Characteritic of the upended ediment regime of the River Creedy. ha been aeed by conidering the variability of load etimate obtained uing replicate data et relating to particular ampling trategie. INTERPOLATION PROCEDURES Interpolation procedure are frequently ued to calculate loading from the regular ampling aociated with water quality urveillance programme, and the accuracy and preciion of a number of thee procedure have been aeed by applying them to replicate et of data repreenting ampling at 1, 2, 4, 7, lo and 14 day interval. Replication ha been undertaken by uing a random number generator to elect ampling time within the

7 The reliability of upended ediment load data 183 interval and retaining thee time throughout the period of record. A tandard 50 replicate ha been generated for each ampling interval, except 1 and 2 day where only 24 and 48 replicate are poible. Sampling on a monthly bai, and involving the collection of 1, 2, and 4 ample per month, ha alo been repreented uing a imilar approach, in order to tet a load calculation procedure requiring etimate of mean monthly concentration. The numerical bae of the ix load calculation procedure employed are lited in Table 1. Method 1 ue the aumption that the value of concentration and dicharge aociated with individual ample may be averaged to provide repreentative mean value for the period of record (e.g. Verhoff et al., 1980). In method 2 the individual value of concentration and dicharge are eentially combined to produce a value of ediment dicharge repreentative of each interval, and ummed over the period of record. Alternatively, thi method can be viewed a calculating the mean of the ampled intantaneou load, which i in turn applied to the whole period of record. Thi approach ha been employed in the UK Harmonized Monitoring Programme (Department of the Environment, 1979). Wherea method 1 and 2 make ue only of the intantaneou flow value aociated with individual ample, the remaining procedure utilize the full flow record from the meauring tation, by incorporating either the mean dicharge for individual interval, mean monthly dicharge value or the mean dicharge for the period of record. Method 3 evaluate load a the product of average concentration and the mean dicharge for the period of record (e.g. Ongley, 1973), whilt method 4 combine the flow-weighted mean concentration with the mean dicharge for the period (e.g. Verhoff et al., 1980). Method 5 aume the ampled concentration i repreentative of the ampling interval and calculate the load a the um of the product of ampled concentration and mean dicharge for individual interval. Finally, method 6 calculate load a the um of mean monthly load value which are in turn calculated a the product of mean monthly ampled concentration and the mean monthly dicharge. The mean monthly concentration may thu be baed on ample from everal year. Thi method wa ued by Ongley et al. (1977). Etimate of the upended ediment load of the River Creedy for the 7 year period, produced uing thee ix interpolation procedure in combination with the data provided by different ampling frequencie, are lited in Table 2. Both the mean and tandard deviation of the replicate etimate are lited. The tandard deviation may be ued a a meaure of the preciion of the etimate, ince it directly reflect the variability of the individual value produced by the replicate ample et. When compared to the actual load for the period of t, the mean value aociated with individual calculation procedure indicate that method 1, 3 and 6 underetimate the load by 70% or more. Thi ugget that the procedure weighting the concentration value by the dicharge at the time of ampling (method 2, 4 and 5) are likely to produce more accurate load etimate.

8 184 D.E. Walling & B.W. Webb Table 1 Load interpolation procedure Method Numerical procedure 1 Total load = K (2 n, A ) (2" ) i-i n i-i n n C i Q i 2 Total load = K2" (-^-) i-i n - Ci 3 Total load = K Q r (Z" ) K2" (CiQi) i-i n Total load = 0, ^Qi 5 Total load = K2[' =1 (CjQpi) 6 Total load = K S ^ (C m Q m ) K = converion factor to take account of period of record Cj = intantaneou concentration aociated with individual ample (mg Qi = intantaneou dicharge at time of ampling (m 3 " ' Q r = mean dicharge for period of record (m 3-1 ) Qpi = mean dicharge for interval between ample (m 3-1 Cm = mean monthly concentration (mg I" 1 ) Qm = mean monthly dicharge (m 3 " 1 ) n = number of ample Table 2 Mean and tandard deviation of replicate upended ediment load etimate for the period obtained uing variou interpolation procedure Load calculation procedure (Table 1 ) Sampling interval (day) X 2 x 3 x 4 x 5 x 6 x x = mean of replicate reult (t) = tandard deviation of replicate reult (t) The tandard deviation value mut, however, alo be conidered and their ignificance in term of preciion i clearly demontrated in Fig. 3. Here, the variability of the etimate, produced by individual interpolation procedure applied to the replicate data et for ampling frequencie of 7 and 14 day, ha been portrayed in an idealized form by plotting

9 The reliability of upended ediment load data 185 Sampling interval - 7 day Supended ediment load (tonne. 10 } Fig. 3 Idealized ditribution of replicate load etimate for the period obtained uing the variou interpolation procedure in combination with data provided by ampling frequencie of 7 and 14 day. Curve number refer to individual interpolation procedure lited in Table 1. the normal ditribution repreented by the appropriate value of mean and tandard deviation. Thee ditribution have been truncated at two tandard deviation and afford an indication of the 95% confidence limit of the replicate load value. Figure 3 exhibit an invere relationhip between accuracy and preciion. Method 2 and 4 produce mean value of load which are cloe to the actual load, but the catter of the replicate etimate i great and it would be extremely difficult to apply a conitent correction factor to load etimated uing thee two method, or to place reliance on the rank order of load calculated for different bain. Method 1 and 3 produce the greatet underetimation of the ediment load for the 7 year period (c. 80%) but provide more conitent reult and may therefore be preferable. The influence of ampling frequency upon the reliability of the reultant load etimate can be aeed by comparing the reult lited in Table 2 for different ampling frequencie. Surpriingly, perhap, ampling frequency appear to have little influence upon the accuracy of method 1-4, a indexed by the mean load value. With method 5, however, the degree of underetimation increae with increaing ampling interval. The influence of ampling interval on preciion, a reflected by the value of tandard deviation, i more marked. In all cae preciion harply decreae with an increaed ampling interval and would appear to be an important criterion in the election of an appropriate frequency. Table 3 introduce a further apect of reliability by conidering a 7 day ampling interval and the ratio of mean load etimated for individual year uing method 1-5 to the actual load for thoe year. The ratio aociated with method 1 and 3 are relatively conitent, again indicating potential for application of a correction factor, whilt thoe produced by method 2 and 4, and to a leer extent 5, exhibit coniderable variability. However, any attempt to make ue of a correction factor in etimating annual load mut alo conider the

10 186 D.E. Walling & B.W.Webb Table 3 A comparion of the ratio between actual load and the mean load obtained for individual year uing the replicate data et, for a 7 day ampling interval and variou interpolation procedure Interpolation procedure Ratio of etimated/meaured load Actual load (t) preciion of the replicate load etimate, ince the ratio lited in Table 3 relate to the mean value for the replicate. A in the cae of the load etimate for the 7 year period, method 1 and 3 are aociated with the lowet value of tandard deviation for the replicate load etimate for individual year. Any attempt to evaluate the relative merit of the load etimation procedure lited in Table 1 mut clearly conider both accuracy and preciion criteria and may involve ome compromie. The low degree of preciion mut effectively rule out method 2 and 4 and the mot worthwhile reult are probably to be obtained from method 1 and 3. Both exhibit a marked tendency to underetimate but the relatively high preciion ugget that the method will reproduce relative ranking and it ha been hown that ome cope exit for the application of correction factor EXTRAPOLATION PROCEDURES The reliability of load etimate obtained uing rating curve technique ha been evaluated in a imilar manner to that employed for interpolation procedure. Four data et, containing 50 replicate and repreenting different ampling trategie, have been aembled from the 7 year record and the reultant rating relationhip have been applied to everal frequently ued load calculation procedure. The ampling trategie employed to generate the replicate data et include regular ampling at weekly interval, and attempt to improve the coverage of torm event by introducing additional random aperiodic ampling when flow exceed certain threhold (Table 4). Fifty replicate rating relationhip, of the form b concentration = aq where Q = intantaneou dicharge at time of ampling, have been etablihed for each of the four ampling trategie, uing leat quare regreion. In addition, the data et for the type 3 and 4 rating were ubdivided according to eaon and to riing and falling tage condition (3a, 4a) and four eparate rating relationhip were developed for each {cf. Walling, 1977a). The

11 The reliability of upended ediment load data 187 Table 4 The ampling trategie ued for rating curve derivation Strategy Sampling programme Regular weekly ampling Strategy 1 plu 200 random ample collected when dicharge > 15 m 3 " 1 Strategy 1 plu 150 random ample collected when dicharge > 15 m 3 " 1 and 50 random ample collected when dicharge > 30 m 3-1 Strategy 1 plu 750 random ample collected when dicharge > 15 m 3 " 1 and 250 random ample collected when dicharge > 30 m 3-1 Type 1 Rating relationhip Type 2 Rating relationhip Ê 1 o,. 1 Type 3 Rating relationhip (eaonal & tage ubdiviion) WR! z T yp e 4 Rating relationhip (eaonal & tage ubdiviipn) "^léfiiii%vc. SR^jlllllv» '. - /^ " -/*Sm&i ^«" '/' Âmt >"V.\ JZ:Ù?:% y ;?>» *)&... WR :,",;,- jifwf 0WF Winter falling El WR Winter riing UJ SF Summer falling 0 SR Summer riing *^ Wi$.»[,&:>. : "/"'T"Û _ "SF - Dicharge (m ) Fig. 4 Example of rating plot and relationhip etablihed uing the four alternative ampling trategie.

12 188 D.E. Walling & B.W.Webb tatitical propertie of the variou rating relationhip are ummarized in Table 5, and Fig. 4 preent example of individual ample et and their aociated rating relationhip. In all cae the catter aociated with the relationhip i coniderable, and the addition of aperiodic torm event ampling can be een to change the lope of the relationhip ignificantly (compare type 1 with type 2). Table 5 Statitical propertie of the rating relationhip a* 4 4a WF WR SF SR WF WR SF SR n r a b r, a, b = mean value of r, a and b for the 50 replicate equation. *Rating ubdivided according to eaon and tage condition a follow: WF = winter and falling tage, WR = winter and riing tage, SF = ummer and falling tage, SR = ummer and riing tage, n value for ubdivided rating are calculated a average value from the 50 replicate data et. A viual indication of the degree of variability within the rating relationhip produced from replicate ample et i afforded by Fig. 5(b). Thi depict a repreentative election of 10 rating for type 1 and type 2 ampling trategie. Although the intercept of the type 1 rating curve are imilar, there i appreciable variation in the exponent or lope. Thi variation i much le for type 2 rating curve and for type 3 and 4. The rating relationhip lited in Table 5 have been applied to the continuou flow erie for the 7 year period of record, in order to calculate the upended ediment load. Following common practice, both hourly and daily mean flow data have been employed, and the reultant load etimate are ummarized in Table 6. In all cae, ue of daily mean flow data reult in etimate which are lower than thoe produced uing the hourly flow erie. Taking the mean of the replicate load etimate a an index of accuracy, it can be een that the ue of thee rating relationhip underetimate the ediment load for the 7 year period by between 83 and 23%. However, the increae in the number of ample, aociated with the progreion from ampling trategy 1 through to trategy 3, i paralleled by an increae in accuracy. Similarly the ue of rating relationhip ubdivided according to eaon and tage tendency produce an increae in the accuracy of the etimate provided by a particular ampling trategy. It i pertinent to note that a previou evaluation of the accuracy of rating curve etimation of ediment load for thi river, undertaken by Walling (1977a), uing a very intenive ampling trategy, indicated that load could be overetimated. Clearly, the ize and repreentativene of the data bae ued to derive the rating relationhip exert an important influence on the

13 The reliability of upended ediment load data 189 Table 6 Mean and tandard deviation of replicate upended ediment load etimate for the period obtained uing variou rating relationhip applied to the hourly and daily mean flow erie Rating relationhip a* 4 4a* Hourly flow erie Daily flow erie X X Not applicable to daily mean flow erie. x = mean of replicate reult (t) = tandard deviation of replicate reult (t) (a) Daily (b)!! '! ' I '!! ' I Supended ediment load (tonne. 10 ) Type 1 rating curve Type 2 rating curve Dicharge (rr? 1 ) Dicharge (rrf ) Fig. 5 (a) Idealized ditribution of replicate load etimate for the period obtained uing the variou rating relationhip in combination with the hourly and daily mean flow erie. Curve number refer to individual rating type, (b) The variability of a repreentative election of replicate rating relationhip of type 1 and type 2. likely accuracy of the reultant load etimate. Figure 5(a) indicate the preciion of the load etimate produced uing the variou rating curve, by depicting the normal ditribution aociated with the relevant value of mean and

14 190 D.E. Walling & B.W.Webb tandard deviation of the replicate load etimate. Preciion tend to increae with the ue of increaing number of ample to derive the rating relationhip. Neverthele, it would appear that ampling trategy 3 produce inufficient data for reliable ubdiviion of the rating relationhip, becaue the pread of load etimate produced by the type 3a rating i coniderable. Overall, the preciion of the rating curve etimate i generally better than that aociated with the interpolation procedure (Table 2). More pecifically, the reult obtained from the variou interpolation procedure applied to data collected at 7 day interval (Fig. 3) may be directly compared with the load etimate producing ued the type 1 rating, alo baed on a regular 7 day ampling trategy (Fig. 5(a)). The relatively high preciion aociated with the rating curve etimate of ediment load for the 7 year period could ugget potential for applying a general correction factor to make allowance for their more limited accuracy. However, Table 7 indicate that uch potential i very retricted. Thi conider the ratio of the mean load, etimated for individual year uing a particular rating, to the actual load for that year. Even with type 4 rating, hown in Fig. 5(a) to produce the highet accuracy and preciion, the ratio varie from 0.19 to 1.36 when uing the hourly flow erie. Thee reult may be compared with thoe preented in Table 3 for the interpolation procedure. In that cae, procedure 3 howed much more conitent ratio, with value lying between 0.17 and To ome extent thi contrat reflect the baic ditinction between interpolation and extrapolation procedure, in that the latter applie data collected from the whole period of record to individual year wherea the former ue only data relating to a particular year. Table 7 A comparion of the ratio between actual load and the mean load for individual year calculated from the replicate etimate produced by individual rating relationhip combined with the hourly flow erie Ratio etimated/meaured load Rating relationhip a 4 4a Actual load (t) Ue of flow duration curve Supended ediment rating curve are alo frequently ued with flow duration curve data to calculate ediment load (e.g. Miller, 1951). Walling (1977b) ha previouly hown how the choice of duration increment can ignificantly influence the magnitude of the reultant load etimate. To explore thi problem further, load etimate for the 7 year period were produced from duration

15 The reliability of upended ediment load data 191 curve of hourly and daily mean flow uing the duration increment uggeted by Miller (1951), Piet (1964), Murthy (1977) and Collin (1970). The reult for type 1 and type 4 replicate rating relationhip are preented in Table 8. The preciion of the duration curve etimate, a indexed by the tandard deviation of the replicate reult, i cloely imilar to that achieved uing the hourly and daily mean flow erie. However, ignificant difference exit between the load etimate produced uing the flow erie and ome of thoe obtained uing duration curve procedure, depite the ue of the ame rating curve. Thi introduce a further element of uncertainty into any aement of the reliability of ediment load data and it ignificance can be expected to increae with decreaing bain ize. Load interval method The "load interval method" i a variant of the rating curve approach which ha been ued by a number of worker in recent year and which mut alo claify a an extrapolation procedure (e.g. Verhoff et al., 1980). In thi method, the dicharge ordinate of the rating plot i partitioned into a number of equal clae and the average load for each cla i calculated a the mean of the load aociated with individual ample falling within that cla. The total load for the period of record i calculated by umming the product of mean load and dicharge frequency for each cla. Thi method ha been applied to the four et of rating curve data, uing dicharge frequency data baed on both the hourly and the daily mean flow erie. The reult are preented in Table 9 and Fig. 6. Interpreting the value of mean and tandard deviation a before, thi method can be een to produce a ignificant improvement in accuracy over the tandard rating curve procedure. Thi improvement may be related to the fact that the mean load aociated with a particular dicharge cla reflect only ample falling in that cla, wherea equivalent etimate baed on rating equation developed uing regreion technique will reflect the trend evidenced by the overall data et. However, thi improvement in accuracy i achieved at the expene of a lo in preciion. The tandard deviation value aociated with the load interval method are everal time greater than thoe produced uing the tandard rating curve technique. The preciion of etimate provided by the data of rating type 1 i unacceptably low, wherea there i little to chooe between the preciion of etimate produced uing data from type 2, 3 and 4 rating. IMPLICATIONS Figure 2 ha hown the coniderable variation in the magnitude of ediment load etimate that may be aociated with indirect method of calculation of long term load uing both interpolation and extrapolation procedure. All the method repreented have been employed frequently by other worker. It i therefore uggeted that detailed attention mut be given to

16 192 D.E. Walling & B.W.Webb Table 8 Mean and tandard deviation of the replicate upended ediment load etimate for the period obtained uing type 1 andtype4 rating relationhip and flow duration curve repreented by different duration increment Rating type Flow duration data Duration increment Miller Piet Murphy Collin 1 4 Hourly Daily Hourly Daily X X X X x = mean of replicate reult (t) = tandard deviation of replicate reult (t) Table 9 Mean and tandard deviation of replicate upended ediment load etimate for the period obtained uing the load interval method Rating data-et Hourly flow erie Daily flow erie X X x = mean of replicate reu It (t) = tandard deviation of replicate reult (t] Load interval calculation Daily flow Supended ediment load (tonne. 10 ) Fig. 6 Idealized ditribution of replicate load etimate for the period obtained uing the "load interval method" in aociation with frequency data baed on the hourly and daily mean flow erie. Curve number refer to individual rating data et.

17 The reliability of upended ediment load data 193 the likely reliability of all uch ediment load data if deign calculation, comparion between bain, and other application are to prove meaningful. Evaluation of the reliability of individual procedure mut involve conideration of both accuracy and preciion and the application of imple correction factor to account for lack of accuracy mut be undertaken with caution. It i advocated that thee problem hould be conidered by the International Standard Organization, WMO and other bodie involved in developing tandard for ediment meaurement. Where reliable method of aeing long term ediment load cannot be applied, the etimate provided by indirect method hould be qualified by a tatement of potential error. For example, the load interval method decribed above ha been developed to produce a tandard error value to qualify the reultant load etimate (Verhoff et al., 1980). Thi tatitic i, however, baed on the tatitical propertie of the individual ample rather than an analyi of the reliability of the method itelf. More work i required on river with different ediment regime, o that we may move toward quantitative aement of the potential reliability of ediment load data and the election of ampling trategie and calculation procedure neceary to achieve required level of accuracy and preciion. ACKNOWLEDGEMENTS The author gratefully acknowledge the upport of the Natural Environment Reearch Council for work on the ediment load of Devon river, the cooperation of the South Wet Water Authority in providing flow data and the aitance of Mr R. Carter with the development of computer analyi. REFERENCES Adam, J. (1980) High ediment yield from major river of the wetern Southern Alp, New Zealand. Nature 287, Campbell, F. B. & Bauder, H. A. (1940) A rating-curve method for determining ilt-dicharge of tream. Tran. AGU 21, Collin, M. B. (1970) Dicuion of "Deign curve for upended load etimation". Proc. Intn Civ. Engr 44, Department of the Environment (1979) Second Biennial Report , Standing Technical Committee on Water Quality. Dept of the Environment, London. Dickinon, W. T., Scott, A. & Wall, G. (1975) Fluvial edimentation in Southern Ontario. Can. J. Earth Sci. 12, Dougla, I. (1971) Comment on the determination of fluvial ediment dicharge. Aut. Geogr. Studie 9, FIASP (1963) Determination of fluvial ediment dicharge. Federal Inter-Agency Sedimentation Project Report no. 14. Fleming, G. (1969) Supended olid monitoring: A comparion between three intrument. Wat. Wat. Engng 72,

18 194 D.E. Walling & B.W.Webb Griffith, G. A. (1979) High ediment yield from major river of the wetern Southern Alp, New Zealand. Nature 282, Loughran, R. J. (1976) The calculation of upended ediment tranport from concentration v. dicharge curve: Chandler River, NSW. Catena 3, Miller, C. R. (1951) Analyi of flow duration ediment rating curve method of computing ediment yield. US Bureau of Reclamation Report. Murthy, B. N. (1977) Life of Reervoir. Central Board of Irrigation and Power, New Delhi, India. Olive, L. J., Rieger, W. A. & Burge, J. S. (1980) Etimation of ediment yield in mall catchment: a geomorphic gueing game? Proc. Conf. Int. Autralian Geographer, Newcatle, NSW. Ongley, E. D. (1973) Sediment dicharge from Canadian bain into Lake Ontario. Can. J. Earth Sci. 12, Ongley, E. D., Ralton, J. G. & Thoma, R. L. (1977) Sediment and nutrient loading to Lake Ontario: methodological argument. Can. J. Earth Sci. 14, Piet, R. F. (1964) Long term ediment yield from mall waterhed. In: Land Eroion, Precipitation, Hydrometry, Soil Moiture (Proc. Berkeley General Aembly of IUGG), IAHS Publ. no. 65. Porterfield, G. (1972) Computation of fluvial-ediment dicharge. Technique of Water Reource Invetigation of the United State Geological Survey, Book 3. Rakoczi, L. (1976) Poibilitie and limitation of nuclear upended ediment gauging. In: Modern Development in Hydrometry, vol. II, WMO no Verhoff, F. H., Yakich, S. M. S Melfi, D. A. (1980) River nutrient and chemical tranport etimation. J. Environ. Engng Div. ASCE 10 (6), Walling, D. E. (1977a) Aeing the accuracy of upended ediment rating curve for a mall bain. Wat. Reour. Re. 13, Walling, D. E. (1977b) Limitation of the rating curve technique for etimating upended ediment load, with particular reference to Britih river. In: Eroion and Solid Matter Tranport in Inland Water (Proc. Pari Symp., July 1977), IAHS Publ. no Walling, D. E. & Kleo, A. H. A. (1979) Sediment yield of river in area of low precipitation: a global view. In: The Hydrology of Area of Low Precipitation (Proc. Canberra Symp., December 1979), IAHS Publ. no Walling, D. E. & Teed, A. (1971) A imple pumping ampler for reearch into upended ediment tranport in mall catchment. J. Hydrol. 13,