We recently published a paper in which, instead of using only rainfall and air temperature (or otehr meteorologica variables) as inout into ANN, we also considered in situ SOIL MOISTURE (see link).

A future interesting research activity is the use of satellite soil moisture data as input, that now are widely and freely available from different sensors.

Good luck,

Luca

Article Fuzzy Logic for Rainfall-Runoff Modelling Considering Soil Moisture

Dear Luca.

Thanks for your reply. I take note for that.

Best wishes,

khairah

In this area I have a book titled: Artificial Neural Network  and GIS for Flood Forecasting,

ISBN-13: 978-3-659-67481-5  

   Published by LAP LAMBERT Academic.

You can use ANN, ANFIS (Hybrid model including ANN and Fuzzy) and many more models but before using those models you should have basic idea about flood forecasting. Flood forecasting is a global as well as local phenomena as it depends on the rainfall intensity, temperature, soil moisture content etc. Most important thing is urbanization (Land Use Land Cover), soil type etc. And all these parameters are not independent . So first you have to check how land use land cover is changing for your study area from the past years. Analyse the trend of precipitation in your area whether it is stationary or non stationary,whether there is any periodicity or not by taking your past data. Plot IDF curve. And if there was flood in previous years in your area then find out the impact of these parameters on the flood. Finally find the most important parameters that is affecting the flood in your area and give as an Input to ANN. But don't forget to do uncertainty analysis after the simulation.

I think this will help you.

You can read the attached paper as well.

Thanks

Dear Colleague

as a first step, you should study the flooding history of the targeted area and focuses on the most hazardous part of the river or stream. The most important parameters in flood modeling are stream and/ or river width, stream depth, stream bathymetry slope, stream meandering, elevation change and rainfall intensity (off course). After that, you can generate several terrain and hydrological parameters such as topographic slope, soil moisture and soil texture ...etc. Those parameters you have to build them using remote sensing data (e.g. DEM, MODIS and PALSAR Data) side by side with field measurements. These measurements will be used to verify the proposed model. Using tree-based models without field observation means nothing. ROC curve can not be used as a strong method to assess the accuracy of the proposed model

Please note that the transitional area between steep and very gentle slope are sites of a very high probability of flooding occurrence

Question: Having identified the contributing factors to flooding comprising of DEM and its derivatives; SPI; TWI; STI; TRI; soil, geology, LULC etc. I would like to use ANN. HOW DOES ONE NORMALIZE THE DATA FOR INPUT INTO THE ANN.

I have followed Saied Pirasteh paper posted above.