I have been working and developing hydrological models in many cases. From the simplest model (Lumped Catchment model) to the integrated and complicated model (MIKE SHE). I realised that every model has its own assumption and limitation. This is something that every hydrologist must understand well. I recommend to all new hydrologist to have a look the manual and standard procedure of any kinds of model do you use. For example, rational method is the simplest and commonly model used to calculate flood discharge. This method is not suitable in a bigger and undeveloped areas. Similarly, physical based model will give you a number of parameters that you need to put. The more detailed the model, the more uncertainty it is. We are dealing with nature which partially cannot be predicted. So know the limitation of the model is important.
The are two ways in hydrological model: stochastic or deterministic. Some scientist may prefer to use deterministic model. This means every steps in hydrological processes are determined. Contrary, stochastic model doesn't really care about the processes. The main concern is inputs and outputs. As long as the input can predict the outputs accurately, then the processes are neglected. We have to understand the limitation of each model. Deterministic models for example, it requires parameters which cannot have an exact values. There are people try to come up with the parameters but all are based on study which may not applicable in other areas. Stochastic model is a future model. Stochastic model relies on samples data. The more we have data the better model prediction. The samples data is also problematic. To predict accurately, the data must be good in quantity and quality. In term of quantity, the numbers of data must really represent the geographical condition spatially. For example, rainfalls depend on latitude, longitude and altitude. These major parameters in determining the rainfall depth. So the numbers of rainfall stations must really represent the latitude, longitude and altitude. In the area where the surface areas are undulating, more stations are required.
 |
| Rainfall Stations |
In term quality, I found many data are not well screened. It was simply collected without any proper screening processes. Field data measurement can be error in many ways. The rainfall data for example, the location and position of the rain gauge is determined the accuracy of the measurement. In monsoon times, the rain was not fell vertically. It was influenced by strong wind which may cause the collected rain are less than the actual rain. If we understand the uncertainty of the data, we can go to next step to develop stochastic model.
I believe, stochastic model is a future model. Many water agencies have invested a lot in data collection. It means in the future, we will have more data accurately. the stochastic model will be much robust then before and it can be used in more wider regions.
 |
| Stochastic Model of Peak Floods |
