Overview of Hydrologic Procedures

1. Chapter 1 of 3. Overview of Hydrologic Procedures

1.1 General

The statistical analysis of observed data or the utilisation of probabilistic data derived by others (e.g. Bureau of Meteorology, Water Authorities etc.) is often the first step taken in hydrologic modelling. The accuracy or reliability of the results produced by hydrologic models is directly dependent on the reliability of the initial input data introduced to the model. The period or length of record available for statistical analysis has a direct influence on the reliability of the analysis. As a general rule probabilistic events of average recurrence intervals up to twice the length of the available data can be reliably estimated from statistical analysis of that data set.

Streamflow and rainfall records are often available to facilitate hydrologic investigations. In situations when there are long streamflow records, an approach to flow estimation based on the statistical properties of the observed streamflow is an attractive option. Methods adopted which utilise the statistical properties of observed streamflow are referred to as streamflow-based methods. For rainfall-based methods requiring the use of a hydrologic model to convert probabilistic rainfall to probabilistic peak flows or flood hydrograph, limits should also be applied to reflect uncertainties in the model parameters.

Instead of limiting the use of these methods, a common approach is to compute confidence limits of flow estimates thus allowing a degree of flexibility and engineering judgement to be applied when designing with these estimates. Hydrologic models should be calibrated either with data available for the specific location of interest or, if these are not available, from nearby catchments. The little data available suggests that an average factor of 0.7 can be used to convert an error in the parameter of a hydrograph model to the resulting error in the peak discharge.

The reliability of any statistical analysis is directly influenced by the number of data available for analysis. In Australia, as in most parts of the world, long streamflow records are less readily available compared to rainfall records. There are only a number of river gauging stations which offer reliable long records of streamflow while there are usually many more rainfall stations which have been in operation for a longer time and thus can offer longer record length. However, the utilisation of these rainfall records for estimation of design discharges require a means of converting the probabilistic design rainfalls to design flows. To facilitate this, many rainfall-based methods of flood estimation have been developed in an attempt to model the hydrological processes necessary to convert the design rainfall to design flows. There are general limits on the applicability of the various flood estimation methods outlined in Table 1.1. These limits relate to inherent assumptions of the methods and the data base from which the parameters of the methods have been derived.

1.2 Selection Of Appropriate Model

The selection of the appropriate method to use in flood estimation is dependent on the requirements of the project and the availability of data. Australian Rainfall and Runoff (Chapter 12) lists eight categories of design activities when providing guidelines in selecting the appropriate method of analysis. These categories are as follows:

​​(i)	​​Routine design of minor to medium works on small to medium size catchments where only the peak discharge is required;
(ii)	Routine design of minor to medium works on small to medium size catchments where a hydrograph is required;
(iii)	Medium to important works, generally on medium to large sized catchment;
(iv)	Extreme floods for major works;
(v)	Diversion or construction floods;
(vi)	Preliminary or check estimates;
(vii)	Deterministic estimates for flood forecasting;
(viii)	Flat areas such as recreation area and airfield.

Table 1.1 overleaf is a summary of applicable methods for the eight activity categories.

1.2.1 Limit on Average Recurrence Interval

The methods used commonly in Australian practice range from the application of the Rational Method to complex Runoff Routing Methods.
As a rough rule:

The Probabilistic Rational Method should not be used for predicting floods of ARI greater than 100 years.

Reliable estimates are limited to ARI equal to twice the second or third longest period of record used in parameter determination.

Reliable estimates from flood frequency analysis are limited to twice the period of observed data available for statistical analysis. Two of these methods will be described in this manual. They are Flood Frequency Analysis and the Index Flood Method.

1.2.2 Choice between Alternative Methods in Ungauged Catchments

A significant number of drainage designs are carried out for sites for which there are no available data for model calibration. In such circumstances, regional methods are used to either estimate probabilistic flows directly or estimate parameter values for application of hydrologic models. Consideration of appropriate regional flood estimation method to adopt include:- Amount of data from the vicinity of the ungauged catchment used in establishing design parameters.

Validity of the physical basis implicit in the available regional method.

Flood Estimation Methods	Design Activities Categories
	(i)	(ii)	(iii)	(iv)	(v)	(vi)	(vii)	(viii)
Arbitrary Methods including Rational, SCS, empirical formulae etc.	Only if no alternatives are available	No	No	No	No	Only if no alternatives are available	Yes	No
Regional Peak Flow Equations	Yes	No	Only as an alternative means of checking results	No	No	Yes>	Only as an alternative means	No
Rational Method	Yes	No	No	No	No	Yes	No	Yes
Probabilistic Rational Method	Yes	No	Yes, if only peak flow is required	No	Yes, if only peak flow is required	Yes	No	No>
Flood Frequency Analysis	Yes	No	Yes, if only peak flow is required	Only as an alternative means of checking results	Yes, if only peak flow is required	Yes	No	No
Design Hydrograph	Not Necessary	Yes	Only as an alternative means	No	No	Not Necessary	No	No
Unit Hydrograph	Not Necessary	Yes	Yes	Yes	Yes	Not Necessary	Yes>	No
Runoff Routing	Not Necessary	Yes	Yes	Yes	Yes	Not Necessary	Yes	No

As a general rule, methods which attempt to model individual processes of flood hydrology are more amenable to regionalisation of its parameters compared to methods which lump all hydrologic processes into a single "black box" model. The order in which flood estimation methods are amenable to regionalisation is as follows:

• Runoff Routing methods
• Probabilistic Rational method
• Regional Flood Frequency method
• Rational method
• Unit Hydrograph method

The potential weaknesses of regional methods are:

(i) not based on sufficient real flood data;

(ii) the model implicit in the method does not validly account for differences in physical and climatic factors between the ungauged catchment of interest and the gauged ones used to establish the model parameters.

The model errors can be minimised by restricting the data used to those from gauged catchments that are hydrologically representative of the ungauged one. A procedure based on a small amount of representative data is likely to be more reliable than one based on a large amount of unrepresentative data; the quantity of data is less important than its representativeness.

1.2.3 Frequency Analysis or Hydrograph Method

It is generally true to say that rainfall records for a particular location will be longer than any runoff records which may exist at the site. They also exhibit less variability. These characteristics are reflected in the frequency curves plotted for observed rainfall and for observed floods. The confidence bands for the rainfall will be much narrower, particularly at high ARI's. The questions that frequently arise are:

(i) for increasing ARI, at what ARI does the flow estimate from a hydrograph method, with its inherent inaccuracies but using more precise rainfall data, become more accurate than values extrapolated from the observed flows.

(ii) what if the frequency curve has been derived from an extrapolated flow record or a regional frequency study.

(iii) if the flood frequency method is the more accurate but a complete hydrograph is required, how can a hydrograph method be adjusted to match the flood frequency method peak.

The approach adopted in ARR is to first classify hydrograph, and other rainfall-based methods in terms of accuracy, as outlined in Table 1.2.

Class	Description	Average Error of estimate
I	Hydrograph methods calibrated "on site"	14%
II	Methods with parameters transposed from nearby catchments, or from regional studies	25-70%
III	Generalised or arbitrary methods (e.g. USDA Soil Conservation Service Method)	> than Class II

A further issue which should be considered in selecting the most appropriate flood estimation method is that pluviograph data are less extensive than general rainfall (daily) data. Consequently, catchments needing short duration data for design (less than 100 km2) are subjected to higher uncertainties as a result of uncertainties due to the design rainfall intensity and storm temporal pattern.

ARR gives equations for the recurrence interval above which hydrograph methods (using rainfall data) are more accurate than frequency analysis of observed streamflows. The answers depend on the class of method, area of catchment, length of flow record, and the variability and skewness of the logarithms of flows.

Reference

Australian Rainfall And Runoff;(1987), A Guide to Flood Estimation, Volume I & II.