Papers and Proceedings of the Royal Society of Tasmania, Volume 121, 1987 23

STREAMFLOW CHARACTERISTICS OF NORTHEASTERN TASMANIA: I. REGIONAL FLOOD FLOWS

by A. D. Knighton

(with four tables and seven text-figures)

KNIG HTON, A.D., 1987 (30:vi): Streamflow characteristics of northeastern Tasmania: 1. Regional flood flows. Pap. Proc. R. Soc. Tasm., 121: 23-33. ISSN 0080-4703. Department of Geography, University of Sheffield, Sheffield, United Kingdom SID 2TN. Based on streamflow records from thirteen stations, regional equations are developed which enable the estimation of mean annual discharge and various flood flows at ungauged sites along rivers in northeastern Tasmania. The area around Swansea is shown to be hydrologically distinct, at least as far as flood discharges are concerned, and the analysis subsequently focuses on the rest of the region. Despite the wide range of climatic and physiographic conditions there, relationships are highly significant, with drainage area explaining more than 97% of the variation in flood discharge. Rates of change are not only relatively high but increase with flood magnitude, suggesting rapid downstream transmission of flood waters. Network magnitude may be a viable alternative to drainage area as an estimator. Its usc has the advantage that the downstream pattern of flow addition can be readily charted for major rivers, as illustrated for the Ringarooma and George, Key Words: Tasmania, stream flow, drainage area, flood discharge.

INTRODUCTION The basic variable used to express the size Macquarie River. Regional How analyses are a and erosive effectiveness of natural rivers is dis­ compromise to a certain extent: between a larger charge, which is generally measured at established sampled area with more gauging stations and a gauging stations. However, the number of mea­ smaller one in which flow conditions are likely to surement sites within an area is usually quite small be more homogeneous. Thus enlargement of the and rarely does a sequence of gauging stations exist sampled area to augment the data base increases along an individual river. Consequently, in order to the degree of heterogeneity and the attendant risk extend a spatially limited data base and provide of sampling from different hydrogeographic popu­ estimates of flow characteristics at ungauged sites, lations. Such proved to be the case here. some form of regional analysis is necessary. The The area can be divided into three main estimates so derived can be used for water manage­ sections. That to the north of the South Esk basin ment and river engineering purposes at regional has rivers draining to the east and north coasts and and basin scales, as well as for routing water and to the Tamar estuary, so that they come under a sediment through individual catchments. Indeed wide range of climatic and physiographic influences. the latter provided the initial stimulus for this study Mean annual precipitation is at a maximum (> 1600 in that reliable flow estimates were required as a mm) to the north and northeast of Ben Nevis, spatial series along the Ringarooma and George where many rivers (notably the North Esk, Rivers in order to model the movement patterns of Ringarooma, George and South Esk) have their material introduced from mining sources. Although headwaters. The first three rivers listed provide a so inspired, the results presented below have impli­ west-east sample across this northern section, the cations for a wider range of purposes. pattern of their mean monthly discharges being The study falls into two parts, the first of distinctively seasonal (fig. 2) and typically temperate which is concerned with flood flow estimation and oceanic (Shaw 1983). Despite the broadly similar the second with hydraulic geometry relationships. regimes with July-September and January-March Both deal with rivers in northeastern Tasmania, a being respectively the maximum and minimum somewhat arbitrarily defined area bounded in the flow periods, small but possibly significant west by a line along the Tamar estuary, South Esk differences are apparent in the monthly runoff and Macquarie Rivers (fig. I). The southern limit is patterns, particularly between the George and the defined by the southern-most sweep of the other two rivers. Even when dealing with averaged 24 A.D. Knighton

Gauging Stations T Tomahawk River GC Goatrock Creek

lC Lau rlston Creek GM Great Musselroe River MA Macquarle River

CU Curries River R RingaroorQa River SE L South Esk, Llewellyn

p Pipers River C Cascade Ri ver SEp South Esk at Perth

arid River G George River NE North Esk B N GF Great Forester River A Apsley River

RA Ransom River S Swan River ME Meredith River t

,,- .... ,, \ " I I I I ~...... ,. .... _'" ".. , I I I I '1.;,1 o 10 20 30 40 50km ....I

FIG. 1 - Distribution of gauging stations in northeastern Tasmania. Streamflow Characteristics of NE Tasmania, 1. Regional Flood Flows 25

flows in neighbouring basins, flow behaviour can RINGAROOMA RIVER 2 be variable. Mean annual The South Esk basin dominates the area with discharge:::; 8'71 m3 5- 1 the main river describing a horseshoe-like course z « (fig. I) along which mean annual precipitation ill :2 varies from 1600 mm in the upper reaches to less than 600 mm south of Perth. The main tributary, r::' o 1 the Macquarie River, has a southern origin and f= « enters the South Esk downstream of Perth, the site a: of the gauging station with the largest drainage area used in this study. In effect the South Esk basin separates the northern section from a southern one represented by a cluster of gauges in the o~------~ neighbourhood of Swansea (fig. 1). There, mean annual precipitation ranges from 650 to 900 mm. GEORGE RIVER

As the following analysis will show, this southern Mean annual z 3 1 section has streamflow properties which are distinct « discharge =6'12 m 5- .----..-.--1 ill from those further north. :2 o f- REGIONAL ANALYSIS g «f­ The estimation of flood frequency or hydro­ a: graph parameters at ungauged sites commonly relies on statistical relationships between river flow and catchment characteristics. The latter can be o - broadly grouped into climatic, surface cover (soils and land use) and drainage basin properties, cate­ I NORTH ESK RIVER gories which include a potentially large number of 2 r Mean annual 3 operational variables. Ease and reliability of ! discharge = 5 43 m s-l measurement, and likely success in prediction are appropriate criteria in the choice of suitable o variables. From the point of view of the first f- criterion, drainage basin properties are probably '2 the most suitable provided reliable topographic f­« maps are available at a large enough scale. Those a: properties can be further subdivided into: (i) size variables - drainage area, mainstream length, total channel length; Oi) slope properties of the channel and hillslope o M A M systems; and A SON 0 (iii) channel network variables drainage density, FIG. 2 - Flow regimes of the Ringarooma, stream frequency, network magnitude; George and North Esk Rivers expressed by the with drainage area and network magnitude being ratio of mean monthly to mean annual discharge. relatively easy to measure. As regards the second criterion, drainage area has consistently been an the most consistent out of 16 independent variables effective variable, one reason for its success being since it was the only one to appear in all ten that it subsumes a great deal of relevant hydro­ analyses. Given also that drainage area has been logical information. In probably the most compre­ successfully used to predict the median flood (Q2) hensive study of streamflow variation yet completed in western Tasmania (Watson 1975, Watson & (NERC, 1975), drainage area proved to be the Williams 1983), its choice for this study seems single most important variable influencing the obvious, especially since this study is largely of an mean annual flood (Q233) at both national and exploratory nature. Following standard practice, regional levels of the British Isles. In assessing ten least-squares regression analysis is used to derive regional flood-frequency analyses from the United the required relationships which are assumed to States, Riggs (1973) found that drainage area was have the typical log-linear form. 26 A.D. Knighton

Mean Annual Discharge achieved by the inclusion of mean annual precipita­ tion as an additional independent variable. Never­ Mean annual discharge is simply the arith­ theless, most of the residuals above the regression metic average of all daily mean flows, with a line have higher than average precipitation, while frequency of 29, 32.5 and 30% respectively at the those below have lower than average. That the two North Esk, Ringarooma and George gauging sta­ South Esk stations plot as negative residuals can be tions. In most humid areas it is closely related to partly attributed to the precipitation effect, drainage area and mean annual precipitation. although channel storage may be a contributory Here, its relationship to drainage area is highly factor at these large drainage areas. However, significant with a coefficient of determination (p2) exclusion of those stations to give of 0.98 and a standard error of estimate of 0.11 02 (fig. 3). Basins of less than 30 km2 rarely figure in Qm. :: 0.01 A/· (I) such plots but there are three in this area and they has obviously little effect on a well-defined regional plot well relative to other points. trend which is maintained over three and a half cycles of logarithms.

oc -----r-r--, 1 ------'1 --~J r- ! Flood Discharges and Drainage Area "S~p SEc The main part of this paper is concerned with

10 R. higher magnitude flows which are more relevant G. hydrologically and morphologically. Flood fre­ NE- .5 quency analysis is a procedure for estimating the GF. P oGM A ~ + peak discharge which on average is likely to be I o 972 < p < 0 997 B. < I I equalled or exceeded once in a specified period of U ' - is 1 T years. Thus QT is the peak flow having a return "I period or recurrence interval of T years. Before proceeding to the regional stage, the magnitudes of I .eu specified flood flows had to be determined for each 01~ OLe of the gauging stations. One basis for flood frequency analysis is the annual maximum series which takes the single peak discharge in each year of record so that the number

00', GC ~L~_~ _ _l~ of data values equals the record length in completed 5 10 50 100 500 1000 DRAINAGE AREA. krnl years. Ten years is generally regarded as the minimum period for analysis and, with one excep­ FIG. 3 - Relationships of mean annual discharge tion, this prescription was followed (table I). to drainage area. The 95% confidence limitsfor the Lauriston Creek was specifically included to extend correlation coefficient are given. the data base into the lower end of the drainage area range where there are few data but, despite Over much of the eastern United States only seven years of record, the additional error where total precipitation varies gradually with involved is probably quite small, particularly in the distance and each unit area contributes about the estimation of average floods. Having ranked the same volume of runoff, mean annual discharge annual peak flows and calculated their recurrence tends to have a one-to-one relationship with drain­ intervals from (n+J)jr (where n is the number of age area (Dunne & Leopold 1978). Such is the case years of record, r is the rank of a particular flood), a here despite the less uniform pattern of rainfall. suitable frequency distribution has to be chosen. Indeed quite steep precipitation gradients exist There is no unequivocal theoretical basis for choos­ from the coastal and South Esk valley areas toward ing the correct distribution to describe an annual the northern highlands. In addition to an exponent maximum series but the log-Pearson Type III is value of almost one, the regression equation (fig. 3) widely accepted (Benson 1968, NER,C 1975) and, in has an intercept value which indicates that on common with Watson (1975), it was used here average a mean annual discharge of 0.01 m3 S-I is because of its objectivity. Generally it provided a produced from a unit basin area. good fit to the data, although several of the plots Such is the dominance of drainage area that had a hint of sinusoidal autocorrelation within the no improvement in the level of explanation was residuals. The graph for the North Esk (fig. 4) Streamflow Characteristics of NE Tasmania, l Regional Flood Flows 27

200 I

'"E w (!l a: ~ 100r------+------~--_,~~L-~--~------~ u en o ~ :::l /.. ~ • X 50 ./' <{ /" /" ~ . ...J ./' ./ <{ :::l T Z ~ Z .. <{ ./ ./

20 / /

1 '11 1'25 2 5 10 50 RECURRENCE INTERVAL.years FIG. 4 - Flood-frequency plotfor the North Esk at Ballroom, 1950-1983. The 95% confidence limits are ~hown. TABLE 1 Flood flow data.

Gauging station Length of Flood discharges, m3 S-1 record, record Index number Rtver years Ql.ll Q2 Qs QlO

302200 Swan 20 97 490 892 1123 302204 Apsley 15 82 188 325 433 302205 George 11 46 116 213 292 302208 Meredith 16 28 97 163 199 302210 Great Musselroe 13 30 70 123 165 302213 Goatrock Creek 10 0.9 1.8 2.9 3.6 318017 South Esk at Llewellyn 25 204 523 971 1341 318204 North Esk 34 29 64 108 142 318208 South Esk at Perth 27 174 491 972 1390 318215 Lauriston Creek 7 2.3 4.1 6.2 7.6 319200 Brid 14 11 25 42 56 319202 Tomahawk 13 15 24 33 40 319204 Pipers 11 23 62 118 165

The index number is the one used by the Rivers and Water Supply Commission of Tasmania. 28 A.D. Knighton

has a recurrence interval of 2.33 years and the ratio ~-T~r of Q2.33/ Q2 at the] 3 gauging stations is fairly stable "SE p with a mean of 1.12 and a standard deviation of xA /',<5 0.02. These average flows are important since it is / believed that discharges in the neighbourhood of GM P 0= 0·99 "c bankfull cumulatively perform most work in natural • NE P rivers and therefore are largely responsible for / forming the channel (Knighton 1984). 84 0 11,= O'23 Ad0 A preliminary analysis in terms of the ratio o 950

0 5 =0 40AdO 98 mania. By way of example, the gauging station on 0950

TABLE 2 Flood flow-drainage area relations, northeastern Tasmania. Flood Equation constants Correlation Standard error discharge a b coefficient of estimate

(i) Northern area - nine stations Q1.l1 0.23 0.84 0.99 0.10 Q2 0.32 0.93 0.99 0.09 Qs 0.40 0.98 0.99 0.10 QlO 0.45 1.01 0.99 0.11 (ii) Southern area - four stations QUI 0.87 0.81 0.99 0.16 Q2 1.62 0.93 0.99 0.03 Q5 2.57 0.95 0.99 0.03 QlO 3.24 0.95 0.99 0.03

Symbols: Q = a Adh.

TABLE 3 Flood flow-drainage area relations. Source Location Drainage Discharge a b area, km2

Benson (1962) New England, USA 4-25070 Q2.JJ 0.56 0.85 164 stations Qs 0.97 0.82 QlO 1.51 0.79 Thomas & Benson Potomac River basin, USA 15-30000 Q2 0.86 0.80 (1970) Qs 1.74 0.78 QlO 2.80 0.76 NERC (1975) British Isles 0.05-9868 Q2.33 0.68 0.77 533 stations Watson (1975) Western Tasmania 449-2539 4.70 0.74 7 stations Watson & Western Tasmania 15-2 539 5.19 0.75 Williams (1983) JO stations

may be due to the low levels of storage available so Tasmania which is wetter, steeper and more densely that translation rather than reservoir effects drained. Those for the southern area are reasonably dominate the downstream transmission of flood high, suggesting a greater capacity for flood flow waters. Comparing tables 2 and 3, it appears that generation within that group of streams. northeastern rivers do indeed increase their flood The amount of scatter about the regression volumes downstream at a faster rate than do rivers lines is small (fig. 5). In the northern area graphs in similar environments. On the other hand, the the George and South Esk (Llewellyn) stations intercept values for the northern area are relatively consistently appear as positive residuals, with the small, particularly when compared with western Brid, North Esk and South Esk (Perth) stations as 30 A.D. Knighton

negative residuals. This variation, although rela­ --I- r-~ tively small, points to inter-basin contrasts in flood generation and transmission. The two South Esk stations seem to be in the wrong order as regards flood volumes (table 1, fig. 5), possibly because water movement to the Perth station is more subject to the influence of temporary storage. However, exclusion of that station from the ,m l Q) 0·41 M regressions had little effect on the form of the odsed Oil 1 100, 000 maps equations which, considering the range of en­ 0-902

vironmental conditions covered, describe the data I very well. based 0" 1: 25,000 maps i 0·734 < p < 0,997 Flood Discharges and Network Topology 51---- LC LC J i X Intuitively it seems reasonable to argue that ~-.~--'----'---- the way in which channels of different size are 10 100 1000 linked together within the drainage network can MAGNITUDE AT CiJl.,UGE influence the movement pattern of flood waters and therefore the resultant flood volumes, especially FIG. 6- Relationship ofthe median annualflood at downstream locations. Certainly the Flood to network magnitude measured at two map scales, Studies Report (NERC 1975) includes stream 1:25000 and 1:10000. frequency (number of channel segments per unit area) as an important explanatory variable of the the data set to six and excludes the upper end of the mean annual flood and attempts have been made to scale represented by the South Esk stations. Based relate parameters of the Instantaneous Unit on measurements of link magnitude (M) and Hydrograph to network variables (Rodriguez­ network diameter(D) at each of the six gauges, the lturbe & Valdes 1979). There, however, the empha­ following equations were obtained for the median sis has been on Horton-Strahler rather than annual flood (Q2) Shreve-Smart measures of drainage network com­ Q2 0.16 M093 (p::: 0.97, SEE 0.13) (3) position. A major element in the more recent = = Shreve-Smart approach is the introduction of new Q2::: 0.33 D 109 (p::: 0.81, SEE::: 0.34) (4) topologic parameters which describe network structure in dimensionless terms. Two are con­ Although significant at the 95% level, the latter sidered here -- magnitude and diameter. The does not describe the distribution of data particu­ magnitude of a channel link, an unbroken section larly well and is henceforth disregarded as a of' channel between successive nodes (sources, possible means of flow estimation. junctions or outlet), is defined as the number of The exponent in equation (3) is exactly the sources upstream, while diameter is the maximum same as in the corresponding drainage area rela­ link distance within the network. Both have geo­ tionship (table 2, fig. 5), which supports the belief metric equivalents - namely, drainage area and that the rate of downstream increase in Q2 along mainstream length. Indeed magnitude (M) and these northern rivers is faster than elsewhere. drainage area (Ad) are related theoretically (Shreve Again the Brid (northern coastal plain) appears as 1967) by a negative resid ual and the George (sharply dissected and rugged headwaters) as a positive one, with the (2) remaining points quite close to the line (fig. 6). For where L is the average link length and K is a comparative purposes and to produce an equation coefficient, but magnitude is easier to measure applicable to those parts of the northern area provided suitable maps are available. outside that currently covered by the 1:25 000 Map scale is known to affect channel delinea­ series, magnitude was determined from 1: 100 000 tion and therefore the calculation of network maps for the original nine northern stations and indices, so that it has become common practice to correlated with the median annual flood to give take the blue-line network shown on 1:25000 maps Q2::: 0.41 M lol (p:: 0.98, SEE::: 0.14) (5) as the initial basis for measurement. Unfortunately three of the nine northern stations are not yet The regression coefficient is slightly larger than in covered by the new 1:25000 series, which reduces equation (3) but the intercept value is significantly Streamflow Characteristics of NE Tasmania, I. Regional Flood Flows 31

TABLE 4 Flood flow~network magnitude relations. Flood Equation constants Correlation Standard error discharge a b coefficient of estimate

0) 1:25 000 series - six stations Ql.lI 0.13 0.82 0.97 0.12 Q2 0.16 0.93 0.97 0.13 Q5 0.19 1.00 0.97 0.15 (ii) 1: 100000 series - nine stations Ql.lI 0.28 0.92 0.98 0.12 Q2 0.41 1.01 0.98 0.14 Q5 0.54 1.06 0.97 0.17

Symbols: Q =aMbo so, indicating that a unit stream at I: 100000 has on plotted for the George and Ringarooma rivers average a magnitude of about 3 at the larger scale. (fig. 7). Scale limitations do not allow the smaller Although equation (5) explains the data quite well tributaries to be shown but there is sufficient detail (fig. 6), it is not considered to be as reliable as to indicate major differences between the two river equation (3) because that is based on 1:25 000 maps systems. Whereas the George is characterised by which are both more recent and better defined as infrequent, large-scale additions of flow, concen­ regards surface channels. trated in the middle section, the Ringarooma has a The fact that magnitude correlates reasonably more regular pattern with smaller but more fre­ well with Q2 (and other flood flows, table 4) implies quent tributary contributions along the entire that it could be an effective alternative to drainage length of the river. In part these differences can be area as an estimator of flood discharge. In choosing associated with the more elongate form of the between the two, likely success in prediction and Ringarooma basin (fig. 1). Such plots emphasize ease of measurement have already been mentioned the discontinuous nature of flow addition, which as appropriate criteria. Judging from the higher has implications for the transport of sediment and correlation and lower standard error of estimate, the adjustment of channel form. They may also drainage area seems to be superior as a predictor provide a basis for more efficient water manage­ but the magnitude relationship, which is based on ment and for routing flood flows through drainage fewer points, could be improved when the entire networks. area is mapped at the 1:25000 scale. As regards the second criterion, the measurement of drainage area CONCLUSION is usually straightforward but can be time consum­ ing if flow is to be estimated at many ungauged Given that hydrologic response is influenced sites. On the other hand, magnitude is easily by many factors which are usually highly correlated measured and, because network structure can be and spatially variable, regional analyses based on simply stored in a computer as a binary string, the equations having only one independent variable magnitude of any point can be readily determined. inevitably contain errors. Consequently flow esti­ The main requirement is consistency in the delinea­ mation must be tinged with caution, especially if tion of networks and their constituent channels. the available data base is relatively small as is the The use of magnitude has a further advantage. case here. Also, application of the results to longer Since each channel link within a network has a time periods may be invalid if, as in New South particular magnitude, the discharge of each link Wales (Pickup 1976), the rivers are subject to can be estimated and the pattern of flow addition fluctuating regimes over 30-50 year periods. Set from tributary sources traced for major rivers. against these reservations is the high level of Based on measurements of network structure and significance achieved by the regional relationships the regional discharge-magnitude relationship for northeastern Tasmania, in which drainage area (equation 3), the downstream pattern has been explains more than 97% of the variation in flood 32 A.D. Knighton

Littlechilds Creek BO R. GEORGE 1 Power Rivulet ,-. 70 1 ME ci 60 0 ~ 50 ;t North George ::J Rivet Z «Z 40 1 z « 0 ~ 30 ~ !;( Beckers Creek ::;; 20 i= Dobsons South George x- - -­ ~ Creek 1 I _.1----x 10 1 _I .... - _ ... "" '--North George

L-.--__~ I ! 0 10 20 30 40 50 60 Pig and Whistle

R. RINGAROOMA Big Valley ~ 140 Creek ,- River 0 Wyniford I Me 120 River c· Weld L 0 J ~ 100 ;t ::J legerwood zZ « 80 Rivulet z Dorset « River J 0 60 ~ Maur'ice 1 River ~ !;( ::;; i= ~ ~I20

0

I I I I I I I ! I I 0 10 20 30 40 50 60 70 80 90 100 110 120

DISTANCE DOWNSTREAM, km

FIG. 7 - The downstream pattern of flow addition estimated from equation (3) for the George and Ringarooma Rivers.

discharges (table 2, fig. 5). Atypically the higher flows more than five times as large as those magnitude floods increase faster with drainage elsewhere in the area. Finer distinctions can be seen area than do the lower magnitude ones, which may in the pattern of residuals from the northern reflect limited storage during the downstream regressions (figs 5 and 6). The George River, which transmission of flood water. The analysis has also drains one of the wettest catchments, and the South shown that the rivers in the neighbourhood of Esk at Llewellyn always appear as positive residuals, Swansea are hydrologically distinct, with flood while the Brid and South Esk at Perth have the Streamflow Characteristics of NE Tasmania, I. Regional Flood Flows 33

opposite quality. The Great M usselroe and particu­ REFERENCES larly the Pipers Rivers show very little deviation from the regression lines and from that point of DENSON, M.A., 1962: Factors influencing the occur­ rence of floods in a humid region of diverse view may be regarded as typical basins worthy of terrain. U.S. G.S. Water-Supply Pap. 1580-B: more intensive study. 64pp. Network magnitude appears to offer a viable BENSON, M.A., 1968: Uniform flood frequency esti­ alternative to drainage area as an estimator of mating methods for federal agencies, Water flood flows in single-variable equations. Its use Resources Research, 4: 891-908, provides an opportunity for charting the pattern of DUNNE, T. & LEOPOLD, L.B., 1978: WATER IN flow addition along major rivers (fig. 7) and may ENVIRONMENTAL PLANNING. W.H. Free­ improve estimation procedures by taking account man, San Francisco. of tributary inflows. However, any further applica­ KNIGHTON, D., 1984: FLUVIAL FORMS AND PROCESSES. Edward Arnold, London. tion must await the completion of the 1:25 000 map NERC, 1975: FLOOD STUDIES REPORT. Natural coverage on which surface channels are better Environment Research Council, London. defined. Indeed the overall need is for more data to PICKUP, G., 1976: Geomorphic effects of changes in improve the reliability of results. The relationships river runoff, Cumberland Basin, N.S.W. Aus­ obtained in this initial study should be re-evaluated tralian Geographer, 13: 188-193. in five years' time when the data base is more RIGGS, H.C., 1973: Regional analyses of streamflow extensive through both an increased length of characteristics. U.S.G.S. Techniques of Water­ record and a larger number of suitable gauging Resources Investigations, Bk 4, Ch B3: 15 pp. stations. RODRIGUEZ-ITURDE, I. & VALDES,J.B., 1979: The geomorphic structure of hydrologic response. Water Resources Research, 15: 1409-1420. ACKNOWLEDGEMENTS SHAW, E.M., 1983: HYDROLOGY IN PRACTICE. Van Nostrand Reinhold (UK), Wokingham. I wish to thank The Royal Society 'and the SHREVE, R.1,., 1967: Infinite topologically random University of Sheffield Research Fund for financial channel networks. J. Geol., 75: 178-187. assistance; Dr Les Wood, the Head of the Depart­ THOMAS, D.M. & BENSON, M.A., 1970: Generaliza­ ment of Geography, University of Tasmania, for tion of streamflow characteristics from drainage­ providing study facilities; Mike Williams of the basin characteristics. U.S.G.S. Water-Supply Hydro-Electric Commission; and particularly Pap. 1975: 55 pp. WATSON, B., 1975: Flood magnitude and frequency Douglas Steane and Andrew Livingston of the Part I: West Coast of Tasmania, Annual Flood Rivers and Water Supply Commission for making Series. Hydro-Electric Commission, Investiga­ their flow records available. tions Division Report: 9 pp. WATSON, B. & WILLIAMS, M.L., 1983: Flood peak estimation for small catchments on the west coast of Tasmania. Hydro-Electric Commission, Investigations Division Report: 6 pp. (accepted 15 September 1986)