usgs.py 文件源码

def avg_wl(self, numObs=50, avgtype='stdWL', grptype='bytime', grper='12M'):
        """Calculates standardized statistics for a list of stations or a huc from the USGS

        :param numObs: minimum observations per site required to include site in analysis; default is 50
        :param avgtype: averaging technique for site data; options are 'avgDiffWL' and 'stdWL'; default is 'stWL'
        :param grptype: way to group the averaged data; options are 'bytime' or 'monthly' or user input; default 'bytime'
        :param grper: only used if 'bytime' called; defaults to '12M'; other times can be put in
        :return:
        """

        data = self.cleanGWL(self.data)
        # stationWL = pd.merge(siteinfo, data, on = 'site_no')
        data.reset_index(inplace=True)
        data.set_index(['datetime'], inplace=True)
        # get averages by year, month, and site number
        site_size = data.groupby('site_no').size()
        wl_long = data[data['site_no'].isin(list(site_size[site_size >= numObs].index.values))]
        siteList = list(wl_long.site_no.unique())
        for site in siteList:
            mean = wl_long.ix[wl_long.site_no == site, 'value'].mean()
            std = wl_long.ix[wl_long.site_no == site, 'value'].std()
            wl_long.ix[wl_long.site_no == site, 'avgDiffWL'] = wl_long.ix[wl_long.site_no == site, 'value'] - mean
            wl_long.ix[wl_long.site_no == site, 'stdWL'] = wl_long.ix[wl_long.site_no == site, 'avgDiffWL'] / std

        if grptype == 'bytime':
            grp = pd.TimeGrouper(grper)
        elif grptype == 'monthly':
            grp = wl_long.index.month
        else:
            grp = grptype
        wl_stats = wl_long.groupby([grp])[avgtype].agg({'mean': np.mean, 'median': np.median,
                                                        'standard': np.std,
                                                        'cnt': (lambda x: np.count_nonzero(~np.isnan(x))),
                                                        'err_pls': (lambda x: np.mean(x) + (np.std(x) * 1.96)),
                                                        'err_min': (lambda x: np.mean(x) - (np.std(x) * 1.96))})

        return wl_stats