Example #1
0
 def get_data(self, dates=None, products=None, chunk=None):
     """ Read all files as time series, stacking all products """
     # TODO - change to absolute dates
     days = numpy.array([int(d.strftime('%j')) for d in dates])
     imgarr = []
     if dates is None:
         dates = self.dates
     if products is None:
         products = self.requested_products
     for p in products:
         gimg = self.get_timeseries(p, dates=dates)
         # TODO - move numpy.squeeze into swig interface file?
         ch = gippy.Recti(chunk[0], chunk[1], chunk[2], chunk[3])
         arr = numpy.squeeze(gimg.TimeSeries(days.astype('float64'), ch))
         arr[arr == gimg[0].NoDataValue()] = numpy.nan
         if len(days) == 1:
             dims = arr.shape
             arr = arr.reshape(1, dims[0], dims[1])
         imgarr.append(arr)
     data = numpy.vstack(tuple(imgarr))
     return data
Example #2
0
    def get_aod(cls, lat, lon, date, fetch=True):
        """Returns an aod value for the given lat/lon.

        If the pixel has a no-data value, nearby values are averaged.  If
        no nearby values are available, it makes an estimate using
        long-term averages.
        """
        pixx = int(numpy.round(float(lon) + 179.5))
        pixy = int(numpy.round(89.5 - float(lat)))
        roi = gippy.Recti(pixx - 1, pixy - 1, 3, 3)
        # try reading actual data file first
        aod = numpy.nan
        with utils.error_handler('Unable to load aod values',
                                 continuable=True):
            # this is just for fetching the data
            inv = cls.inventory(dates=date.strftime('%Y-%j'),
                                fetch=fetch,
                                products=['aod'])
            img = inv[date].tiles[cls.Asset.Repository._the_tile].open('aod')
            vals = img[0].Read(roi)
            # TODO - do this automagically in swig wrapper
            vals[numpy.where(vals == img[0].NoDataValue())] = numpy.nan
            aod = vals[1, 1]
            img = None
            source = 'MODIS (MOD08_D3)'
            # if invalid center but valid vals exist in 3x3
            if numpy.isnan(aod) and numpy.any(~numpy.isnan(vals)):
                aod = numpy.mean(vals[~numpy.isnan(vals)])
                source = 'MODIS (MOD08_D3) spatial average'

        # Calculate best estimate from multiple sources
        if numpy.isnan(aod):
            day = date.strftime('%j')
            repo = cls.Asset.Repository
            cpath = repo.path('composites')
            nodata = -32768

            source = 'Weighted estimate using MODIS LTA values'

            def _calculate_estimate(filename):
                val, var = cls._read_point(filename, roi, nodata)
                aod = numpy.nan
                norm = numpy.nan

                # Negative values don't make sense
                if val < 0:
                    val = 0

                if var == 0:
                    # There is only one observation, so make up
                    # the variance.
                    if val == 0:
                        var = 0.15
                    else:
                        var = val / 2

                if not numpy.isnan(val) and not numpy.isnan(var):
                    aod = val / var
                    norm = 1.0 / var
                    utils.verbose_out('AOD: LTA-Daily = %s, %s' % (val, var),
                                      3)

                return aod, norm

            # LTA-Daily
            filename = os.path.join(cpath, 'ltad',
                                    'ltad%s.tif' % str(day).zfill(4))
            daily_aod, daily_norm = _calculate_estimate(filename)

            # LTA
            lta_aod, lta_norm = _calculate_estimate(
                os.path.join(cpath, 'lta.tif'))

            if numpy.isnan(lta_aod):
                raise Exception("Could not retrieve AOD")

            aod = lta_aod
            norm = lta_norm
            if not numpy.isnan(daily_aod):
                aod = aod + daily_aod
                norm = norm + daily_norm

            # TODO - adjacent days

            # Final AOD estimate
            aod = aod / norm

        utils.verbose_out('AOD: Source = %s Value = %s' % (source, aod), 2)
        return (source, aod)
Example #3
0
    def get_aod(cls, lat, lon, date, fetch=True):
        pixx = int(numpy.round(float(lon) + 179.5))
        pixy = int(numpy.round(89.5 - float(lat)))
        roi = gippy.Recti(pixx - 1, pixy - 1, 3, 3)
        # try reading actual data file first
        try:
            # this is just for fetching the data
            inv = cls.inventory(dates=date.strftime('%Y-%j'),
                                fetch=fetch,
                                products=['aod'])
            img = inv[date].tiles[''].open('aod')
            vals = img[0].Read(roi)
            # TODO - do this automagically in swig wrapper
            vals[numpy.where(vals == img[0].NoDataValue())] = numpy.nan
            aod = vals[1, 1]
            img = None
            source = 'MODIS (MOD08_D3)'
            # if invalid center but valid vals exist in 3x3
            if numpy.isnan(aod) and numpy.any(~numpy.isnan(vals)):
                aod = numpy.mean(vals[~numpy.isnan(vals)])
                source = 'MODIS (MOD08_D3) spatial average'
        except Exception:
            VerboseOut(traceback.format_exc(), 4)
            aod = numpy.nan

        var = 0
        totalvar = 0

        day = date.strftime('%j')
        # Calculate best estimate from multiple sources
        repo = cls.Asset.Repository
        cpath = repo.path('composites')
        if numpy.isnan(aod):
            aod = 0.0
            norm = 0.0
            cnt = 0
            nodata = -32768

            source = 'Weighted estimate using MODIS LTA values'
            # LTA-Daily
            filename = os.path.join(cpath, 'ltad',
                                    'ltad%s.tif' % str(day).zfill(3))
            val, var = cls._read_point(filename, roi, nodata)
            var = var if var != 0.0 else val
            if not numpy.isnan(val) and not numpy.isnan(var):
                aod = val / var
                totalvar = var
                norm = 1.0 / var
                cnt = cnt + 1
                VerboseOut('AOD: LTA-Daily = %s, %s' % (val, var), 3)

            # LTA
            val, var = cls._read_point(os.path.join(cpath, 'lta.tif'), roi,
                                       nodata)
            var = var if var != 0.0 else val
            if not numpy.isnan(val) and not numpy.isnan(var):
                aod = aod + val / var
                totalvar = totalvar + var
                norm = norm + 1.0 / var
                cnt = cnt + 1
                VerboseOut('AOD: LTA = %s, %s' % (val, var), 3)

            # TODO - adjacent days

            # Final AOD estimate
            aod = aod / norm
            totalvar = totalvar / cnt

        if numpy.isnan(aod):
            raise Exception("Could not retrieve AOD")

        VerboseOut('AOD: Source = %s Value = %s' % (source, aod), 2)
        return (source, aod)