Check need for interpolation to compare results

[barneydobson]

Description

@cheginit I've pulled out a sample synthetic results timeseries and compared with a real one. We see that the time indices perfectly align less than 1% of the time, so I think we are going to have to interpolate to make NSE comparisons.

Fixes #76

[cheginit]

Since I cannot reproduce this, I cannot get into specifics. But I am guessing that you're running SWMM at variable time steps and its output does not match the observed streamflow. I haven't used SWMM a lot, but I am wondering if you can set up reporting time at intervals in a such a way that resampling gets you exactly the same date as of the observations. For example, if the observation is hourly, you would report at, for example, 15 min, and just do df.resample("1H").mean() to get the result at the observation dates. This way, you let SWMM take care of setting time steps correctly without the need for interpolation during the post-processing.

[barneydobson]

I've had a look at the option to use swmm_stride which lets you control the timesteps more carefully in pyswmm, but, while it changes the reporting to be precise and uniform, it seems to produce confusing simulation datapoints (at least in my initial test). As far as I can tell it is more intended for custom control structures and not simulation reporting. I will leave this for now - though in checking it has revealed a bug in the interpolation code, so still helpful ;). Possibly when swmmanywhere is made open source, if we explore some collaboration with the pyswmm people I am sure they will have a better idea for how to pre-specify reporting timesteps.

[barneydobson]

Closing this as this branch is way out of date - will reopen the issue if anyone wants to discuss #262