Dear Praveen,

In addition to the good points of Minjiao several other issues related to event-based and continuous rainfall-runoff (RR) modelling.

Event-based RR modelling is often used for short-term purposes such as flood forecasting and flow forecasting in general. Continuous RR modelling is useful for long-term purposes such as the estimation of design discharges, impacts of climate and land use change and impacts of water management strategies.

In event-based RR modelling, the model is run for a number of selected events (e.g. flood events, typically days to weeks) and in continuous RR modelling the model is run for a long time period depending on the data availability and purpose of the study (typically several years to several decades).

Indeed, the estimation of initial state variables is very important for event-based RR modelling and for instance satellite-based information on soil moisture can be very useful. In continuous RR modelling, often a spin-up (warm-up) period is used before the actual calibration is carried out to enable the state variables to adjust to the current hydrological conditions.

Regards,

Martijn

Dear Praveen,

The first thing is event based analysis needs to set initial values for the model statement variables. Wrong initial values may cause very large uncertainties. Sometimes, it will result in over-fitting, in other word fake agreement.

Another one may be event based analysis usually focuses on flood events, and more attention is paid to high flow rate. The agreement of slow runoff processes such as groundwater flow may not be carefully considered.

Regards,

Minjiao

Dear Praveen,

In addition to the good points of Minjiao several other issues related to event-based and continuous rainfall-runoff (RR) modelling.

Event-based RR modelling is often used for short-term purposes such as flood forecasting and flow forecasting in general. Continuous RR modelling is useful for long-term purposes such as the estimation of design discharges, impacts of climate and land use change and impacts of water management strategies.

In event-based RR modelling, the model is run for a number of selected events (e.g. flood events, typically days to weeks) and in continuous RR modelling the model is run for a long time period depending on the data availability and purpose of the study (typically several years to several decades).

Indeed, the estimation of initial state variables is very important for event-based RR modelling and for instance satellite-based information on soil moisture can be very useful. In continuous RR modelling, often a spin-up (warm-up) period is used before the actual calibration is carried out to enable the state variables to adjust to the current hydrological conditions.

Regards,

Martijn

Thank you very much sir @ Minjiao Lu and sir @ Martijn J. Booij for your valuable comments

Oops, I forgot the spin-up issue. As Martijn mentioned, often a spin-up (warm-up) period is used to avoid the effects from initial values in continuous RR modelling. Model outputs in this period are excluded from model evaluation because they could depend on initial values. However, the determination of the spin-up period is not well understood. If it is too long, we lose some information in the data set. If too short, the model output will be affected by the initial values.

Our following paper tried to study how long the spin-up period should be set to.

@Article{Rahman2015Model,

author = {Rahman, Mohammad Mahfuzur and Lu, Minjiao},

title = {Model Spin-Up Behavior for Wet and Dry Basins: A Case Study Using the Xinanjiang Model},

journal = {Water},

volume = {7},

number = {8},

pages = {4256},

year = {2015},

issn = {2073-4441},

doi = {10.3390/w7084256},

url = {http://www.mdpi.com/2073-4441/7/8/4256},

file = {Rahman2015Model.pdf:2015-2019\\Rahman2015Model.pdf:PDF},

owner = {lu},

timestamp = {2015.08.10},

}

As Martijn stated event based RR models are usually applied for example, for infrastructure designing, where we need a storm design to achieve the design hydrograph, usually for maximum and extreme events (and at hourly or less time step). In this cases, one of the main problems of the event based event models is to estimate the initial conditions, for example the antecedent soil moisture. You can use continuous modelling for this purpose. In this case the antecedent soil moisture is estimated for the model. The problem with this models is that you should model the rainfall period and the dry period (continuous modelling). Also, the continuous RR models are more difficult to calibrate (warm up, etc, etc) and if you simulate at a small time step over a long period most of the model are high compute effort demanding. However, if your want to evaluate a water resources system for allocate water resources, in general, continuous models are used but at monthly time step.

I suggest you to read the following papers that deal with this problem, both for maximum design and water resources management design:

1) Gabriel-Martin I., Sordo-Ward A., Garrote L., Castillo L.G., 2017. Influence of initial reservoir level and gate failure in dam safety analysis. Stochastic approach. Journal of Hydrology, 550, 669–684. doi:10.1016/j.jhydrol.2017.05.032.

2) Sordo-Ward A., Granados I., Martín-Carrasco, F. Garrote L., 2016. Impact of Hydrological Uncertainty on Water Management Decisions. Water Resources Management, 30: 5535, doi:10.1007/s11269-016-1505-5. Q1

3) Sordo-Ward A., Bianucci P., Garrote L. and Granados A., 2016. The influence of the annual number of storms on the derivation of the flood frequency curve through event-based simulation. Water, 8, 335, doi:10.3390/w8080335. Q2

4) Flores-Montoya I., Sordo-Ward A., Mediero L., Garrote L., 2016. Fully Stochastic Distributed Methodology for Multivariate Flood Frequency. Water, 8, 225, 1-20, doi:10.3390/w8060225, Q2

5) Sordo-Ward A., Bianucci, P., Garrote L., Granados, A., 2014. How safe is hydrologic infrastructure design? Analysis of factors affecting extreme flood estimation. Journal of Hydrologic Engineering. doi: 10.1061/(ASCE)HE.1943-5584.0000981, Q1

6) Sordo-Ward A., Garrote L., Bejarano M.D, Castillo L.G., 2013. Extreme flood abatement in large dams with gate-controlled spillways. Journal of Hydrology, 498(2013), 113–123, DOI: 10.1016/j.jhydrol.2013.06.010. Q1.

7) Bianucci P., Sordo-Ward A., Pérez J.I., Mediero L., Garrote L. and García-Palacios J., 2013. Probabilistic Calibration of the Parameters of a Reservoir Flood Control Model by Using Streamflow Ensemble. Natural Hazards and Earth System Sciences, 13(4), 965-981, 2013. DOI: 10.5194/nhess-13-1-2013. Q2

8) Sordo-Ward A., Garrote L., Martín-Carrasco F. and Bejarano M.D, 2012. Extreme flood abatement in large dams with fixed-crest spillways. Journal of Hydrology, 466-467, 60-72. DOI:10.1016/j.jhydrol.2012.08.009. Q1

Dear Praveen,

If I was to summarize with only two words, the main objectives of event-based and continuous RR modeling, I would say this:

Event-based RR modeling: flood forecast (flood risk assessment)

Continuous RR modeling: future trend

All colleagues have provided very valuable and valid information on the subject. Your question has been more than adequately answered. Allow me, though, to slightly contribute to the discussion by the following points:

The continuous simulation of the hydrological processes can be an indicator of the future trend concerning their quantification (surface runoff, baseflow, direct runoff, total discharge).

The deviations between computed and measured water discharge values for single rainfall events can be mitigated by means of the continuous RR modeling, because of the integrating effect obtained through the use of a long simulation period.

Through the continuous RR modeling for a relatively long time period in a relatively large basin, a more realistic representation of the runoff processes can be achieved. Additionally, the maximum peak values of the hydrographs might be easier to be computed.

Very important for a continuous hydrologic model is the simulation of time variation of baseflow during and after the rainfall events. In event-based simulations the variation with time of baseflow is insignificant and thus, it does not need to be estimated; a fixed value of baseflow could be considered.

Generally, in event-based RR models, the rainfall losses due mainly to infiltration are computed. However, in continuous RR models, the rainfall losses due to evapotranspiration are mainly computed because they constitute the largest part of the rainfall losses on a long-term time basis.

Kind Regards,

Kostas

P.S.: All the aforementioned are mentioned in the following article and thesis, in which you can find more detailed information.

Kaffas K. and Hrissanthou V., 2014. Application of a Continuous Rainfall-Runoff Model to the Basin of Kosynthos River Using the Hydrologic Software HEC-HMS, Global Nest Journal, 16(1): 188-203.

Kaffas K., 2017. Development of Mathematical Model for Calculating Continuous Hydrographs and Sediment Graphs in a Basin Due to Rainfall. Ph.D. Dissertation, Democritus University of Thrace, Greece, 153 pp.

Hi all

All answers are very useful and reasonable.

the best

azizian

Please share me the best answer might you get...