Alternating Block Method for Stochastic Rainfall

Purpose

This script aims to reduce the time spent converting Rainfall-Intensity Duration Frequency Curves (RIDF) into usable rainfall data for Hydrologic and Hydraulic Simulations.

Project Author

• Ned Santiago, started on November 28, 2022

Usage

Change directory to the source directory

cd ./src/

Run main.py

py ./main.py

Select a csv file with the following format, where each value uses mm/hr as its units:

	10	20	30	60	120	180	360	720	1440
2	123.9	96.1	81.3	55.8	37.5	28.6	18.0	11.0	6.3
5	171.9	132.5	111.8	76.0	51.8	40.2	25.9	16.1	9.5
10	203.7	156.5	131.9	89.4	61.3	47.9	31.1	19.5	11.5
20	234.2	179.6	151.2	102.3	70.4	55.3	36.1	22.7	13.5
25	243.8	186.9	157.4	106.4	73.3	57.6	37.7	23.7	14.1
50	273.6	209.5	176.3	118.9	82.2	64.8	42.5	26.9	16.0
100	303.2	231.9	195.0	131.4	91.0	72.0	47.4	30.0	17.9

Select where to save the result

Components

• Data Handler
  • DirectoryHandler - retrieves the file paths and stores for use
• Rainfall Calculator
  • rainfall curve fit function
  • Ridf - an ridf object that holds the rainfal-intensity duration table
  • AlterBlock - rearranges the time series data into an alternating block result
• curve fit
  • CurveFitter - the calculator for curve fitting
• visualize results
  • Grapher - graphs and produces figures

Future Updates

• Github integration
• Pytest integration
• Singleton data handler, separate data handler and path taker
• Provide a GUI to the DataRetriever object
• Make a small website to act as an improved GUI implementation
• Create a test for each class
• Organize classes into clearly defined python files
• Logging file that shows the process of the calculation, important for debugging the software
• Use class handlers.RotatingFileHandler in log file to also force log file size limit
• Change Ridf class
  • Ridf should enforce that durations should be in minutes format, always increasing from left to right
  • Maybe enforce that it should be an intensity scale i.e. mm/hr per return-period-duration and not mm per return-period-duration
  • Ridf should act more like a dataframe with functionality to check data validity. Thus, need to inherit pd.Dataframe
  • Ridf can have an abstract method so that it can be built on by another class
• Change CurveFitter class
  • CurveFitter should immediately start curve fitting after receiving the path to the data -> CurveFitter runs the curve fit algorithm after receiving a dataframe
  • Optimize the CurveFitter code, many variables were not used or not needed (e.g. some column and index length variables)
  • CurveFitter curvefit method should only work on one row of data, another method should be responsible for compiling to table -> this is already the functionality of scipy curve_fit
  • CurveFitter curvefit method should work with any formula or varying variable counts of formulas 'abc', 'abcd', 'a', etc. -> CurveFitter now has this functionality
  • CurveFitter's estimate_data method should convert an appended list into a dataframe and stop using concat dataframe
  • CurveFitter now accepts list of x values for input, making this a more scalable solution
• Possible Data Interfacer between other objects and Grapher object to separate duties of data transformation and display, takes a dataframe and converts it into sets of x and y data for a grapher. This could be in the Grapher object itself.
• Change AlterBlock class
  • AlterBlock can be a dataframe with added functionality to create alternating blocks.
  • AlterBlock should rewrite the axis with possibility for time-series-like indexes
• Remove (redundant) DataRecorder class since new AlterBlock class will inherit the needed functionality pandas dataframe class
• Setup a new test that uses the old method as the test data, difference should be less than 5%
• Use minutes as the fundamental data type of the analysis, so that everything can be an integer. (Was forced to re-convert back to hours because using minutes added instability to the curve fit calculation)
• Variable time steps
• Add error handling
• Selection of rainfall stations
• Use as a plugin to other programs
• CurveFitter detects how many parameter a function requires and makes a list of columns accordingly
• Make a documentation, possible logo of droplet hole in a graph
• Maybe add this to a personal website as part of a portfolio
• Curvefitter should detect whether the requested return period or duration already exists in the dataframe
• CurveFitter should be able to take multiple return period or duration requests
• Adding a function that properly formats dataframes to correct progression of return period and durations.
• rainfall_curve_fit should be more flexible with what time-series data it can estimate (1hr-, 30min-, 1min-interval)
• remove xdata and ydata functionality form Ridf since the dataframe attributes handle this already.
• remove func argument in graph_data and estimate_data because it should always be using the private _formula attribute