STORM 2

STORM 2 (STOchastic Rainstorm Model) is an upgrade on STORM 1.0, continuing to be a simple and flexible rainfall generator.

FEATURES

• Simulates (circular) rainfall storms from a set of PDFs (Probability Density Functions).
• PDFs easily defined by the user (or retrieved from gauge data).
• Shapefile (SHP) of the catchment required.
• Provided a Digital Elevation Model (DEM), simulates rainfall storms with regard to elevation.
• Spatial resolution (and CRS - Coordinate Reference System) customizable.
• Spatial operations under a raster framework, which adds speed, versatility, and scalability.
• Optimal output storing in NetCDF format.
• Allows rainfall simulation under future (and very likely) climate scenarios.

Improvements over STORM 1.0

• Modelling of storm intensity and duration follows a (bi-variate) Gaussian copula framework.
• Modelling of such Intensity-Duration copulas can be applied at different elevation bands/ranges within the catchment, to account for orographic characteristics influencing precipitation.
• Updated radial decay-rate model for rainfall intensity to incorporate potential, but unrecorded, maximum storm intensities.
• Modelling of storm starting day-of-year (DOY) and time-of-day (TOD) follows a Circular Statistics approach (i.e., mixture of von Mises PDFs).
• Pre-processing module to construct all necessary PDFs from gauge data.
• Output compressed into (geo-referenced) NetCDF files, readily available for visualization (e.g., via Panoply).

HOW TO RUN STORM 2

To smoothly run STORM 2 you'd essentially need two components/files: PDFs and PARAMETERS.

PDFs

The majority of the stochastic component of STORM 2 is located in the file: Probabilty Desity Functions. For every season, and for each variable to be modelled (or sampled from), this file contains the name of the distribution (following scipy nomenclature), and its (nameless) parameters.

There are roughly three ways to generate this file:

• Manual: You can directly (over-)write this (or this) PDF-files with whatever parameters (you think) suit your data the best. For instance, if you want to model Total Seasonal Rainfall TOTALP for Season 1 (in the logarithmic-space) as a left-skewed Gumbel distribution with $\mu = 5.51$, and $\sigma^2 = 0.266$, write TOTALP_PDF1+gumbel_l,5.51,0.226. If you want to model, for Season 2 (and for a -previously defined- elevation band "Z3"), Maximum Rainfall Intensity (at the storm's centre) MAXINT as a generalized Pareto distribution with $\mu = 0.1655$, $\sigma^2 = 0.45431$, and $skewness = 6.427$, write MAXINT_PDF2+Z3+genpareto,0.1655,0.45431,6.427, for instance.

• Automatic: You can use the pre_processing script. If you have gauge data stored/post-processed in this way (for storm rainfall events), and this way (for monthly aggregated rainfall), the script seeks through the data, fits the best PDF for each variable (for all the requested seasons), and generates the aforementioned file of PDFs.
  

• Hybrid: If you have knowledge/idea of the PDFs that describe the storm statistics (see Manual), you can still use the pre_processing script to pass these parameters as (Python) dictionaries.

For each season you want to model/simulate/validate in STORM 2, you'd need at least PDFs for the variables TOTALP - Total Seasonal Rainfall, RADIUS - Maximum Storm Radius, BETPAR - Decay Rate (of maximum rainfall from the storm's centre towards its maximum radius), MAXINT - Maximum Rainfall Intensity, AVGDUR - (Average) Storm Duration, COPULA - Copula's Correlation parameter, and DOYEAR - Storm's Starting Date. DATIME - Storm's Starting Time is optional.

PARAMETERS

The file parameters hosts the main file-paths, variables, and constants which define the storm's modelling characteristics, along with potential climate scenarios, and all the spatio(-temporal) attributes of the catchment.

It is divided into two sections:

• SOFT parameters: These parameters define the type of run (SIMulation/VALidation) - MODE, the number of seasons (1 or 2 -or even more!, potentially) - SEASONS, the number of SIMulations/VALidations per season - NUMSIMS, the number of years to SIMulate/VALidate - NUMSIMYRS, and scaling factors -PTOT_SC, PTOT_SF, STORMINESS_SC, STORMINESS_SF- to (steadily OR progressively) increase/decrease the variables TOTALP and/or MAXINT in order to represent (future) climatic changing conditions. These parameters can be either defined in this file or passed/modified directly from the command prompt (or console) when running STORM 2.

• HARD parameters: These parameters define the location of all input/output files, the elevation bands for which modelling of storm rainfall could be split into, the resolution and CRS of the spatial domain, and some time-domain specifications. These parameters cannot be passed from the command prompt.

RUNNING IT

You can run STORM 2 in its default settings (see PARAMETERS) by calling the storm script as:

python storm.py     # (from your Terminal -and inside your CONDA environment, if so)
%%python storm.py   # (from your Python console)

Type python storm.py -h to prompt what parameters (along with their short explanation) can be passed as optional arguments:

(py39) path-to\STORM>python storm.py -h
usage: storm.py [-h] [-m {simulation,validation}] [-w {1,2}] [-n NUMSIMS] [-y NUMSIMYRS]
                [-ps PTOT_SC [PTOT_SC ...]] [-pf PTOT_SF [PTOT_SF ...]]
                [-ss STORMINESS_SC [STORMINESS_SC ...]] [-sf STORMINESS_SF [STORMINESS_SF ...]] [--version]

STOchastic Rainstorm Model [STORM v2.0]

optional arguments:
  -h, --help            show this help message and exit
  -m {simulation,validation}, --MODE {simulation,validation}
                        Type of Run (case-insensitive) (default: SImuLAtiON)
  -w {1,2}, --SEASONS {1,2}
                        Number of Seasons (per Run) (default: 1)
  -n NUMSIMS, --NUMSIMS NUMSIMS
                        Number of runs per Season (default: 2)
  -y NUMSIMYRS, --NUMSIMYRS NUMSIMYRS
                        Number of years per run (per Season) (default: 3)
  -ps PTOT_SC [PTOT_SC ...], --PTOT_SC PTOT_SC [PTOT_SC ...]
                        Relative change in the seasonal rain equally applied to every simulated year. (one
                        signed scalar per Season).
  -pf PTOT_SF [PTOT_SF ...], --PTOT_SF PTOT_SF [PTOT_SF ...]
                        Relative change in the seasonal rain progressively applied to every simulated year.
                        (one signed scalar per Season).
  -ss STORMINESS_SC [STORMINESS_SC ...], --STORMINESS_SC STORMINESS_SC [STORMINESS_SC ...]
                        Relative change in the observed intensity equally applied to every simulated year.
                        (one signed scalar per Season).
  -sf STORMINESS_SF [STORMINESS_SF ...], --STORMINESS_SF STORMINESS_SF [STORMINESS_SF ...]
                        Relative change in the observed intensity progressively applied to every simulated
                        year. (one signed scalar per Season).
  --version             show program's version number and exit

how does it look like?

The GIF above is produced by the animation script.

Authored and maintained by Manuel F.

GitHub @feliperiosg  ·  Twitter @feLlx