spurious lufft precipitation rates

[jasonebox]

I've started working with the lufft precipitation rates :-)

two related issues...

1.) instead of posting accumulated precipitation, should instead convert precip_u, precip_u_cor (and in some cases precip_l, precip_l_cor) to the time rate of change using e.g.,

            flag=0
            if sum(df.columns=='t_l')==0:
                varnams=['precip_u','precip_u_cor','t_u']
            else:
                varnams=['precip_l','precip_l_cor','t_l','precip_u','precip_u_cor','t_u']
                flag=1

...

            x=pd.read_csv(tmpfile)
            N=len(x)
            if N>0:
                x['tp_u']=np.nan
                for i in range(1,N):
                    x['tp_u'][i]=x['precip_u_cor'][i]-x['precip_u_cor'][i-1]
                    if x['tp_u'][i]>400 or x['tp_u'][i]<0:
                        issues_site.append(site)
                        issues_date.append(df.time[i])
                        issues_p_rate.append(x['tp_u'][i])
                        issues_level.append('u')
                        # x['tp_u'][i]=np.nan

...By posting the rate of change, we have no issue with defining a) when zero should start and b) in cases of extended gaps, accumulated precipitation becomes ambiguous. Rather, users should just see NaN and thus know no data be available that period. Further, users can produce accumulated precipitation, if they wish.

2.) in some cases, get nice results like:

in other cases, the result reflects large offsets in the accumulated precipitation record:

or

there are several other cases...

So, I coded this (from which the example code above comes), that outputs the dates of spurious accumulated precipitation change into this folder (all from this repo). My hope is the issues can be resolved and the next release will have a much more user friendly form. :-)

[jasonebox]

when there are short gaps from no transmission, I think spline interpolation is worth a try. See hourly example below.