Documente Academic
Documente Profesional
Documente Cultură
341
[4]
RAINFALL-RUNOFF
FUTURE
E. TODINI
University of Bologna
(Received January, 1988; revised and accepted February 4, 1988)
ABSTRACT
Todini, E., 1988. Rainfall runoff modeling - - Past, present and future. J. Hydrol., 100: 341-352.
A brief review of the historical development of mathematical methods used in rainfall runoff
modeling is presented. A simple classification of the current available models based upon both a
priori knowledge and problem requirements is proposed in order to assess the state of the art.
Finally an analysis of emerging problems in hydrology is used to ascertain possible future developments and trends.
INTRODUCTION
T h e a n a l y s i s of t h e h i s t o r i c a l d e v e l o p m e n t of a b r a n c h of science s h o u l d
a l w a y s be p r e c e d e d by a d i s c u s s i o n of the r e a s o n s w h i c h in the first place
p r o m o t e d , a n d c o n s t a n t l y m o t i v a t e d , t h e i n t e r e s t in t h a t p a r t i c u l a r line of
research.
T h e s e m o t i v a t i o n s c o n s t i t u t e the b a s i c k e y to u n d e r s t a n d p a s t d e v e l o p m e n t s
a n d f u t u r e t r e n d s in a p a r t i c u l a r b r a n c h of science.
In this p a p e r a n a t t e m p t is m a d e to p r o v i d e a link b e t w e e n the p r o b l e m s t h a t
o r i g i n a t e d a n d m o t i v a t e d the i n t e r e s t of h y d r o l o g i s t s in t h e field of m a t h e m a t i cal i n t e r p r e t a t i o n of the r a i n f a l l - r u n o f f process, a small b r a n c h on t h e t r e e of
h y d r o l o g y . H i s t o r i c a l d e v e l o p m e n t s a n d f u t u r e t r e n d s of r a i n f a l l - r u n o f f
m o d e l l i n g are c o n s i d e r e d in t h e l i g h t of t h e e v o l u t i o n of p r o b l e m s a n d s o l u t i o n s
at d i f f e r e n t p o i n t s in time, w i t h i n t h e f r a m e w o r k of a v a i l a b l e k n o w l e d g e a n d
computational resources.
A HISTORICAL OVERVIEW OF METHODS
T h e origins of r a i n f a l l - r u n o f f m o d e l i n g in the b r o a d sense c a n be found in
the s e c o n d h a l f of the 19th c e n t u r y , a r i s i n g in r e s p o n s e to t h r e e t y p e s of
e n g i n e e r i n g problems: u r b a n s e w e r design, l a n d r e c l a m a t i o n d r a i n a g e s y s t e m s
design a n d r e s e r v o i r s p i l l w a y design.
In all t h e s e p r o b l e m s the design d i s c h a r g e w a s the m a j o r p a r a m e t e r of
interest.
342
According to Dooge (1957, 1973), during the last part of the 19th century and
the earlier part of the 20th, most engineers used either empirical formulas,
derived for particular cases and applied to other cases under the assumption
that conditions were similar enough, or the "rational method" which may be
seen as the first attempt to approach rationally the problem of predicting runoff
from rainfall.
The method, which was derived for small or mountainous catchments was
based upon the concept of concentration time; the maximum discharge, caused
by a given rainfall intensity, happens when rainfall duration equals or is larger
t h a n the concentration time.
During the 1920s, when the need for a corresponding formula for larger
catchments was perceived, many modifications were introduced in the rational
method in order to cope with the nonuniform distribution, in space and time,
of rainfall and catchment characteristics.
The modified rational method, based on the concept of isochrones, or lines
of equal travel time, can be seen as the first basic rainfall runoff model based
on a transfer function, whose shape and parameters were derived by means of
topographic maps and the use of Mannings formula to evaluate the different
travel times.
The types of problems to be solved were much the same as before, but
hydrologists were trying to provide more realistic and accurate solutions,
although still in terms of surface runoff.
Sherman (1932) introduced the concept of the unit hydrograph on the basis
of the superposition principle.
Although not yet known at the time, the superposition principle implied
many assumptions, i.e. the catchment behaves like a causative, linear time
invariant system with respect to the rainfall/surface runoff transformation.
The unit hydrograph principle accelerated the interest of hydrologists who
were now in a position to produce estimators, not only of the peak discharge,
but also of the hydrographs caused by more complex storms.
The unit hydrograph method had however a number of problems: (1) the
separation of surface runoff from base flow; (2) the effective rainfall determination; and (3) the derivation of the unit hydrograph.
Solutions to all these problems involved an extremely high degree of subjectivity.
At the end of the thirties and during the forties a number of techniques were
proposed in order to improve the objectivity of methods and results, and the
techniques of statistical analysis were invoked.
A discussion on the different approaches and the relevant bibliography can
be found in an interesting report by Dooge (1973).
The real breakthrough came in the fifties when hydrologists became aware
of system engineering approaches used for the analysis of complex dynamic
systems. They finally realised that the unit hydrograph was the solution of a
causative, linear time invariant system and that the use of mathematical
techniques such as Z, Laplace or Fourier transforms could lead to the
derivation of the response function from the analysis of input and output data.
34:~
This was the period when "conceptual" models originated. The derivation of
the unit hydrograph in discretized form (the unit graph) from sampled data
(known as the inverse problem) still remained a big problem, due to the nonparticularly linear behaviour of the system and the generally large errors in
input and output data.
To overcome this problem, hydrologists found that shapes of the unit
hydrograph could be provided on the basis of the solution of more or less
simplified differential equations, such as for instance those describing the time
behaviour of the storage in a reservoir or in a cascade of reservoirs (Nash, 1958,
1960). The unit hydrograph could then be expressed in terms of few parameters
to be estimated from catchment characteristics or by means of statistical
procedures: moments, cumulants, regression, maximum likelihood, etc.
A bloom of these model gave rise an unbelievable variety of solutions: a
cascade of linear reservoirs, linear channels, linear channels and reservoirs,
nonlinear reservoirs (Prasad, 1967), etc.
On the other hand, in deriving the unit graph shape from actual data, very
few advances were made even using the transforms: a classical paper by Rao
and Delleur (1971) shows the effect of noise in data, on the unit hydrograph
derived by Fourier transforms. Only after the work of Wiener (1949) and
Tikhonov (1963a, b) and the introduction of continuity and regularization
constraints in the estimation phase [Eagleson et al. (1965), Natale and Todini
(1977)] more realistic and reliable estimates of the unit hydrograph were
obtained.
Studies for the representation of nonlinear systems where also carried out
by means of Volterra integrals and polynominal projection (Amorocho and
Orlob, 1961) or constrained piecewise linearization (Todini and Wallis, 1977).
Box and Jenkins (1970) provided hydrologists with an alternative method of
expressing the unit graph in terms of parameters, i.e. the autoregressive
exogenous variables form of the transfer function (ARX) or the autoregressive
moving average (ARMA), and the analogies with existing "conceptual" models
were pointed out (Spolia and Chander, 1974).
In subsequent developments these techniques, though satisfactory from the
mathematical and philosophical point of view, lost more and more of their
connection with the real world of hydrological problems, and became more or
less mathematical games played by algebrists concerned only to prove the
generality of their approach.
There is an obvious danger in models which fail to provide a reflection of
reality: it is very easy to check a model on the basis of actual rainfall and runoff
values, but it is extremely difficult to understand the quality of results e.g., in
terms of continuity of mass, for a model which uses transformed input and
output in the Box-Jenkins (1970) sense, in order to stationarize the time series
(i.e. through successive applications of the difference operator).
In the sixties other approaches to rainfall-runoff modelling were considered.
In search for a more physical interpretation of the process one could represent
the behaviour of single components of the hydrolog cycle, at the catchment
344
345
The seventies saw a growing awareness and concern for soil erosion and
degradation problems, pollutants diffusion, and in general of the environmental impact of anthropization and land use changes.
Another type of rainfall-runoff model was developed in the late seventies
and in the eighties: the real-time forecasting model as an answer for the need
of warnings in flood prone areas, and as a tool for reservoirs or hydraulic
structures management.
Generally based on recent updating and recalibrating techniques such as
Kalman filters (Kalman, 1960; Kalman and Bucy, 1961; Todini, 1978; Todini and
Wallis, 1978; Kitanidis and Bras, 1980a, b; O'Connell, 1980; Wood, 1980; Wood
and O'Connell, 1985), real-time forecasting models must be reliable, updatable
and constitute part of an entire forecasting system which must include
automatic real-time data acquisition, data validation, forecasting procedures
and dissemination of forecasts (Nemec, 1986; Bacchi et al. 1986).
Although many rainfall-runoff conceptual models have been set in a realtime forecasting mode (recently an intercomparison of models was carried out
in Vancouver on behalf of WMO (1988), there is not yet a clear understanding
on the advantages (or disadvantages) of recalibrating in real time all model
parameters.
AN ASSESSMENT OF PRESENT
In order to find a path among the plethora of different rainfall-runoff
mathematical models available today it seems necessary to introduce an
additional classification of models (which unfortunately will also increase the
plethora of different classifications).
A mathematical model in broad sense, is a combination of two basic
components. The first one expresses all the a priori knowledge that one has on
the phenomenon to be represented and can be referred to as the physical
component.
The second, the stochastic component, expresses in statistical terms what
cannot be explained by the degree of a priori knowledge already introduced in
the physical component (see also Clarke, 1973).
The a priori knowledge can be introduced in the models in many different
ways, ranging from the total a priori ignorance, thus reducing the model to a
pure stochastic process model where not even the cause and effect postulate is
advocated, to the full description of system dynamics based upon the partial
differential equations describing the balance of mass and momentum.
The stochastic component is then added to explain what is not explained by
the physical component, and therefore becomes conditional upon it.
The level of a priori knowledge on the system under study is here taken as
the basis for model classification both in terms of model structure and in terms
of parameters.
On these grounds it is reasonable to assume four classes of model structures
with increasing level of a priori knowledge: (1) purely stochastic; (2) lumped
346
integral; (3) d i s t r i b u t e d integral; and (4) d i s t r i b u t e d differential. Models 2, 3,
and 4 are causal models. Two classes of p a r a m e t e r s can be c o n s i d e r e d as a
f u n c t i o n of m e t h o d s used for t h e i r estimation: (a) stochastic; and (b) physical.
As m e n t i o n e d earlier a p a r t from the p u r e l y s t o c h a s t i c models one can
always add to the causal models a s t o c h a s t i c c o m p o n e n t to explain as m u c h as
possible the u n c e r t a i n t y c o n t e n t of model residuals. In synthesis, the a priori
k n o w l e d g e based classification of models can be r e p r e s e n t e d by Table 1.
Purely stochastic models
= g(x,~,~)
S.
L.
I.
D.
I.
D,
D.
S.
P.
S, = stochastic; L.I. = lumped integral; D.I. = distributed integral; D,D. = distributed differential; P. = physical.
347
348
Finally the distributed differential type, of which very few examples are
known (e.g. SHE), represents the catchment behaviour in terms of all the
differential equations discretized in time and space, expressing mass and
momentum balance for each subsystem and linking together the subsystems by
matching at each step in time their mutual boundary conditions.
The difficulty in the integration and the high computer requirements,
together with the need for large amounts of data has reduced, for the time
being, the applicability of this type of model, although interesting perspectives
can be foreseen in future for its ability in actually describing in a distributed
manner the internal phenomena at the subcatchment scale. This may allow for,
e.g., the study of internal transport and diffusion phenomena as well as the
analysis of the effect of averaging and lumping hydrological quantities or
model parameters at a subcatchment scale in order to obtain more realistic
lumped integral or distributed integral models.
Parameter classes
= g'(M, ~, x, e)
349
In other words, if one knows a priori that the system is well represented by
model M and parameters ~, one should capitalize on that, thus eliminating a
great deal of uncertainty inherent in any statistical analysis.
To clarify this classification on parameter estimates, the parabolic-type
response function model, which is generally used for propagating flow overland
and in channels can be considered a stochastic-parameter lumped integral
model if its parameters (diffusivity and convectivity coefficients) are
determined on the basis of residuals minimizations, or a physically-based
parameter lumped integral model if the parameters are derived on the basis of
their physical interpretation.
TOWARDS FUTURE DEVELOPMENTS
350
REFERENCES
Abbott, M.B., Bathurst, J.C., Cunge, J.A., O'Connell, P.E. and Rasmussen, J., 1986a. An introduction to the European Hydrological System Syst~me Hydrologique Europ~en, "SHE", 1:
History and philosophy of a physically-based, distributed modelling system. J. Hydrol., 87:
45 -59.
Abbott, M.B., Bathurst, J.C., Cunge, J.A., O'Connell, P.E. and Rasmussen, J., 1986b. An introduction to the European Hydrological System Syst~me Hydrologique Europ~en, "SHE", 2:
Structure of physically-based, distributed modelling system. J. Hydrol., 87:61 77.
Amorocho, J. and Orlob, G.T., 1961. Nonlinear analysis of hydrologic systems. Water Resources
Centre, Contrib. No. 40, University'of California, Berkeley, Calif.
Bacchi, B., Burlando, P., Rosso, R. and Todini, E., 1986. Hydrological models for real time flood
forecasting. Proc. Int. Conf. Arno Proj., Florence CNR-GNDCI, Rep. 29.
Beven K., 1975. Distributed Models In: M.G. Anderson and T.P. Burt (Editors), Hydrological
Forecasting. Wiley, New York, N.Y.
Box, G.E.P. and Jenkins, G.M., 1970. Time Series Analysis Forecasting and Control. Holden Day,
San Francisco, Calif.
Burnash, R.J.C., Ferral, R.L. and McGuire, R.A., 1973. A general streamflow simulation system
Conceptual modelling for digital computers. Report by the Joint Federal State River Forecasts
Center, Sacramento, Calif.
Clarke, R.T., 1973. Mathematical models in Hydrology. Irrig. Drain. Pap., 19, FAO. Rome.
Chiu, C.L. (Editor), 1978. AGU Chapman Conference on Application of Kalman Filtering Theory
and Techniques in Hydrology. Hydraul. Water Resour. Dept. Civ. Eng., University of
Pittsburgh, Pittsburg, Pa,
Crawford, N.H. and Linsley, R.K., 1966. Digital simulation in hydrology, Stanford watershed model
IV. Tech. Rep. No. 39, Dep. Civ. Eng., Stanford University.
Dawdy, D.R. and O'Donnel, T., 1965. Mathematical models of catchment behaviour. J. Hydraul.
Div., Proc. ASCE, HY4.
Dooge, J.C.I., 1957. The Rational Method for Estimating Flood Peak. Engineering (London), 184:
311 3374.
Dooge, J.C.I., 1959. A general theory of the unit hydrograph. J. Geophys. Res., 64(2): 241 256.
Dooge, J.C.I., 1973. The linear theory of hydrologic systems. Tech. Bull. U.S. Dep. Agric., No. 1468,
U.S. Gov. Print. Off., Washington, D.C.
Eagleson, P.S,, Mejia, R. and March, F., 1965. The computation of optimum realizable unit
hydrographs from rainfall and runoff data. Hydrodyn. Lab. Rep., No. 84, MIT.
351
Fleming, G., 1975. Computer Simulation Techniques in Hydrology, Environmental Science Series.
Elsevier, Amsterdam.
Freeze, R.A. and Witherspoon, P.A., 1966-1968. Theoretical analysis of regional groundwater flow,
Part I, II, III. Water Resour. Res., Vol. 2(4); Vol. 3(2); Vol. 4(3).
Hebson, C. and Wood, E.F., 1983. Adaptive parameter estimation and filtering of partitioned
hydrologic systems, Dep. Civ. Eng., Princeton University, Princeton, N.J.
IAHS, 1980. Symposium on Hydrological Forecasting. IAHS/AISH Publ. 129.
Kalmam R.E., 1960. A new approach to linear filtering and prediction problems. J. Basic Eng.
(ASME), 82D: 35 45.
Kalmam R.E. and Bucy, R., 1961. New results in linear filtering and prediction. J. Basic Eng.
(ASME), 83D: 9~108.
Kibler, D.F. and Woolhiser, D.A., 1970. The kinematic cascade as a hydrologic model. Hydrol. Pap.
No. 39, Colorado State University, Colo.
Kitanidis, P.K. and Bras, R.L., 1980a. Real-time forecasting with a conceptual hydrologic model
1. Analysis of uncertainty. Water Resour. Res., 16 (6): 1025 1033.
Kitanidis, P.K. and Bras, R.L., 1980b. Real-time forecasting with a conceptual hydrologic model,
2. Application and results. Water Resour. Res., 16 (6): 1034- 1044.
Kulandaiswamy, V.C., 1964. A basic study of rainfall excess surface runoff relationship in a basin
system. Ph.D Thesis, Dep. Civ. Eng., University of Illinois.
Moore, R.J. and O'Connell, P.E., 1978. Real time forecasting of flood event using a transfer function
noise model. Rep. Inst. Hydrol., Wallingford.
Moore, R.J. and Clarke, R.T., 1981. A distribution function approach to rainfall runoff modelling.
Water Resour. Res., 17 (5): 1367 1382.
Nash, J.E., 1958. The form of the instantaneous unit hydrograph. IUGG Gen. Assem. Toronto, Vol.
III, Publ. No. 45, IAHS, Gentbrugge, pp. 114- 121.
Nash, J.E., 1960. A unit hydrograph study, with particular reference to British catchments. Proc.
Inst. Civ. Eng.
Natale, L. and Todini, E., 1977. A constrained parameter estimation technique for linear models in
hydrology. In: T.A. Ciridni et al. (Editors), Mathematical Models for Surface Water Hydrology.
Wiley, New York, N.Y.
Nemec, J., 1986. Hydrological Forecasting. Reidel, Dordrecht.
O'Connell, P.E., 1980. Real time hydrological forecasting and control. Rep. Inst. Hydrol., Wallingford.
Peck, E., 1976. Catchment modelling and initial parameter estimation for the National Weather
Service River Forecast System. NOAA Tech. Memo., NWS Hydro-31.
Prasad, R., 1967. A non linear hydrologic system response model. J. Hydraul. Div.. A~CE HY4.
Rao, R.A. and Delleur. J.W., 1971. The instantaneous unit hydrograph: its calculation by the
transform method and noise control by digital filtering. Purdue Univ. Water Resour. Res.
Centre, Tech. Rep. No. 20.
Rocwood, D.M. and Nelson, M.L., 1966. Computer application to streamflow synthesis and
reservoir regulation. IV Int. Conf. Irrig. Drain.
Sherman, L.K., 1932. Streamflow from rainfall by the unit graph method. Eng. News Rec., 108:
501 5O5.
Simpson, J.R., T.R. and Hamlin, M.J., 1980. Simple self correcting models for forecasting flows on
small basins in real-time. Proc. Oxford Symp. Hydrol. Forecasting, IAHR Publ. 129: 433444.
Sittner, W.T., Schaus, C.E. and Monro, J.C., 1969. Continuous hydrograph synthesis with an
API-type hydrologic model. Water Resour. Res. (5): 1007 1022.
Spolia, S.K. and Chander S., 1974. Modelling of surface runoff systems by an ARMA model. J.
Hydrol., 317 332.
Tikhonov, A.N., 1963a. Solution of incorrectly formulated problems and the regularization method.
Soy. Math., 4: 103~1038.
Tikhonov, A.N., 1963b. Regularization of incorrectly posed problems. Sov. Math., 4:1624 1627.
Todini, E., 1978. Mutually Interactive State P a r a m e t e r (MISP) estimation. Proc. AGU Chapman
Conf. Appl. Kalman Filtering Tech. Hydrol., Hydraul., Water Resour., Dep. Civ. Eng.
University of Pittsburg, Pittsburg, Pa.
352
Todini, E. and Wallis, J.R., 1977. Using CLS for daily or longer period rainfall runoff modelling.
In: Mathematical Models for Surface Water Hydrology. Wiley, New York, N.Y.
Todini, E. and Wallis, J.R., 1978. A real-time rainfall-runoff model for an on-line flood warning
system. Proc. AGU Chapman Conf. Appl. Kalman Filtering Tech. Hydrol., Hydraul., Water
Resour., Dep. Civ. Eng., University of Pittsburg, Pittsburg, Pa.
Wiener, N., 1949. The Extrapolation, Interpolation and Smoothing of Stationary time series. Wiley.
New York, N.Y.
WMO, 1975. [ntercomparison of Conceptual models used in operational hydrological forecasting.
Oper. Hydrol. Rep. 7, WMO No. 429, Geneva.
WMO, 1983. Guide to Hydrological Practices. WMO Publ., No. 168. 4th ed., Geneva.
WMO, 1988. Simulated in Real-Time Intercomparison of Hydrological Models. WMO Publ., WMO,
Geneva (in press).
Wood, E.F., 1980. Recent Developments in Real-Time Forecasting/Control of Water Resources
Systems. Pergamon, Oxford.
Wood. E.F. and O'Connell, P.E., 1985. Real-Time Forecasting. In: M.G. Anderson and T.P. Burt
(Editors), Hydrological Forecasting.
Wooding, R.A., 196~1966. A hydraulic model for the catchment-stream problem, 1. Kinematic wave
theory; 2. Numerical solutions; 3. Comparison with runoff observations. J. Hydrol., Vol. 3,4.
Woolhiser, D.A., Hanson C.L. and Kuhlman, A.R.. 1970. Overland flow on rangeland watersheds.
J. Hydrol., 336~356.
Woolhiser, D.A. and Liggett, J.A., 1967. Unsteady one-dimensional flow over a plane the rising
hydrograph. Water Resour. Res., 3(3): 753-771.
Zhao, R.J., 1977. Flood forecasting method for humid regions of China. East China College of
Hydraulic Engineering, Nanjing, China.