Mercurial > public > atmospheric_lidar / file revision

# General imports
import datetime
from operator import itemgetter

# Science imports
import numpy as np
import matplotlib as mpl
from matplotlib import pyplot as plt
import netCDF4 as netcdf

netcdf_format = 'NETCDF3_CLASSIC' # choose one of 'NETCDF3_CLASSIC', 'NETCDF3_64BIT', 'NETCDF4_CLASSIC' and 'NETCDF4'


class BaseLidarMeasurement():
    """ This is the general measurement object.
    It is meant to become a general measurement object 
    independent of the input files.
    
    Each subclass should implement the following:
    * the import_file method.
    * set the "extra_netcdf_parameters" variable to a dictionary that includes the appropriate parameters.
    
    You can override the get_PT method to define a custom procedure to get ground temperature and pressure.
    The one implemented by default is by using the MILOS meteorological station data. 
    
    """
    
    def __init__(self, filelist= None):
        self.info = {}
        self.dimensions = {}
        self.variables =  {}
        self.channels = {}
        self.attributes = {}
        self.files = []
        self.dark_measurement = None
        if filelist:
            self.import_files(filelist)
        
    def import_files(self,filelist):
        for f in filelist:
            self.import_file(f)
        self.update()

    def import_file(self,filename):
        raise NotImplementedError('Importing files should be defined in the instrument-specific subclass.')

    def update(self):
        '''
        Update the the info, variables and dimensions of the lidar measurement based 
        on the information found in the channels.
        
        Reading of the scan_angles parameter is not implemented.
        '''
        
        # Initialize
        start_time =[]
        stop_time = []
        points = []
        all_time_scales = []
        channel_name_list = []
        
        # Get the information from all the channels
        for channel_name, channel in self.channels.items():
            channel.update()
            start_time.append(channel.start_time)
            stop_time.append(channel.stop_time)
            points.append(channel.points)
            all_time_scales.append(channel.time)
            channel_name_list.append(channel_name)
            
        # Find the unique time scales used in the channels
        time_scales = set(all_time_scales)
        
        # Update the info dictionary
        self.info['start_time'] = min(start_time)
        self.info['stop_time'] = max(stop_time)
        self.info['duration'] = self.info['stop_time'] - self.info['start_time']
        
        # Update the dimensions dictionary
        self.dimensions['points'] = max(points)
        self.dimensions['channels'] = len(self.channels)
        # self.dimensions['scan angles'] = 1
        self.dimensions['nb_of_time_scales'] = len(time_scales)
        
        # Update the variables dictionary
        # Write time scales in seconds
        raw_Data_Start_Time = []
        raw_Data_Stop_Time = []
        
        for current_time_scale in list(time_scales): 
            raw_start_time = np.array(current_time_scale) - min(start_time) # Time since start_time
            raw_start_in_seconds = np.array([t.seconds for t in raw_start_time]) # Convert in seconds
            raw_Data_Start_Time.append(raw_start_in_seconds) # And add to the list
            # Check if this time scale has measurements every 30 or 60 seconds.
            
            duration = self._get_duration(raw_start_in_seconds)
            
            raw_stop_in_seconds = raw_start_in_seconds + duration
            raw_Data_Stop_Time.append(raw_stop_in_seconds)
            
        self.variables['Raw_Data_Start_Time']= raw_Data_Start_Time
        self.variables['Raw_Data_Stop_Time']= raw_Data_Stop_Time
    
        # Make a dictionary to match time scales and channels
        channel_timescales = []
        for (channel_name, current_time_scale) in zip(channel_name_list, all_time_scales):
            # The following lines are PEARL specific. The reason they are here is not clear.
            # if channel_name =='1064BLR':
            #     channel_name = '1064'
            for (ts,n) in zip(time_scales, range(len(time_scales))):
                if current_time_scale == ts:
                    channel_timescales.append([channel_name,n])
        self.variables['id_timescale'] = dict(channel_timescales)
    
    def _get_duration(self, raw_start_in_seconds):
        ''' Return the duration for a given time scale. In some files (ex. Licel) this
        can be specified from the files themselves. In others this must be guessed.
         
        '''
        # The old method, kept here for reference
        #dt = np.mean(np.diff(raw_start_in_seconds))
        #for d in duration_list:
        #    if abs(dt - d) <15: #If the difference of measuremetns is 10s near the(30 or 60) seconds
        #        duration = d
        
        duration = np.diff(raw_start_in_seconds)[0]

        return duration
    
    def subset_by_channels(self, channel_subset):
        ''' Get a measurement object containing only the channels with names
        contained in the channel_sublet list '''
        
        m = self.__class__() # Create an object of the same type as this one.
        m.channels = dict([(channel, self.channels[channel]) for channel 
                            in channel_subset])
        m.update()
        return m
        
    def subset_by_time(self, start_time, stop_time):
    
        if start_time > stop_time:
            raise ValueError('Stop time should be after start time')
            
        if (start_time < self.info['start_time']) or (stop_time > self.info['stop_time']):
            raise ValueError('The time interval specified is not part of the measurement')

        m = self.__class__() # Create an object of the same type as this one.
        for (channel_name, channel)  in self.channels.items():
            m.channels[channel_name] = channel.subset_by_time(start_time, stop_time)
        m.update()
        return m
    
    def r_plot(self):
        #Make a basic plot of the data.
        #Should include some dictionary with params to make plot stable.
        pass 
    
    def r_pdf(self):
        # Create a pdf report using a basic plot and meta-data.
        pass
    
    def save(self):
        #Save the current state of the object to continue the analysis later.
        pass
        
    def get_PT(self):
        ''' Sets the pressure and temperature at station level .
        The results are stored in the info dictionary.        
        '''
    
        self.info['Temperature'] = 10.0
        self.info['Pressure'] = 930.0
        
    def subtract_dark(self):
        
        if not self.dark_measurement:
            raise IOError('No dark measurements have been imported yet.')
        
        for (channel_name, dark_channel) in self.dark_measurement.channels.iteritems():
            dark_profile = dark_channel.average_profile()
            channel = self.channels[channel_name]
            
            for measurement_time, data in channel.data.iteritems():
                channel.data[measurement_time] = data - dark_profile
                
            channel.update()
    
    def save_as_netcdf(self, filename):
        """Saves the measurement in the netcdf format as required by the SCC.
        Input: filename
        """
        params = self.extra_netcdf_parameters
        needed_parameters = ['Measurement_ID', 'Temperature', 'Pressure']
        
        for parameter in needed_parameters:
            stored_value = self.info.get(parameter, None)
            if stored_value is None:
                raise ValueError('A value needs to be specified for %s' % parameter)
        
                
        dimensions = {'points': 1,
                'channels': 1,
                'time': None,
                'nb_of_time_scales': 1,
                'scan_angles': 1,}  # Mandatory dimensions. Time bck not implemented

        global_att = {'Measurement_ID': None,
                    'RawData_Start_Date': None,
                    'RawData_Start_Time_UT': None,
                    'RawData_Stop_Time_UT': None,
                    'RawBck_Start_Date': None,
                    'RawBck_Start_Time_UT': None,
                    'RawBck_Stop_Time_UT': None,
                    'Sounding_File_Name': None,
                    'LR_File_Name': None,
                    'Overlap_File_Name': None,
                    'Location': None,
                    'System': None,
                    'Latitude_degrees_north': None,
                    'Longitude_degrees_east': None,
                    'Altitude_meter_asl': None}

        channel_variables = \
                    {'channel_ID': (('channels', ), 'i'),
                    'Background_Low': (('channels', ), 'd'),
                    'Background_High': (('channels', ), 'd'),
                    'LR_Input': (('channels', ), 'i'),
                    'DAQ_Range': (('channels', ), 'd'), 
                    'Depolarization_Factor': (('channels', ), 'd'), }
                    
        
        channels = self.channels.keys()

        input_values = dict(self.dimensions, **self.variables)
        
        # Add some mandatory global attributes
        input_values['Measurement_ID'] = self.info['Measurement_ID']
        input_values['RawData_Start_Date'] = '\'%s\'' % self.info['start_time'].strftime('%Y%m%d')
        input_values['RawData_Start_Time_UT'] = '\'%s\'' % self.info['start_time'].strftime('%H%M%S')
        input_values['RawData_Stop_Time_UT'] = '\'%s\'' % self.info['stop_time'].strftime('%H%M%S')
        
        # Add some optional global attributes
        input_values['System'] = params.general_parameters['System']
        input_values['Latitude_degrees_north'] = params.general_parameters['Latitude_degrees_north']
        input_values['Longitude_degrees_east'] = params.general_parameters['Longitude_degrees_east']
        input_values['Altitude_meter_asl'] = params.general_parameters['Altitude_meter_asl']
        
        # Open a netCDF4 file
        f = netcdf.Dataset(filename,'w', format = netcdf_format) # the format is specified in the begining of the file
        
        # Create the dimensions in the file
        for (d,v) in dimensions.iteritems():
            v = input_values.pop(d, v)
            f.createDimension(d,v)
        
        # Create global attributes
        for (attrib,value) in global_att.iteritems():
            val = input_values.pop(attrib,value)
            if val: 
                exec('f.%s = %s' % (attrib,val))
        
        """ Variables """
        # Write the values of fixes channel parameters
        for (var,t) in channel_variables.iteritems():
            temp_v = f.createVariable(var,t[1],t[0])
            for (channel, n) in zip(channels, range(len(channels))):
                temp_v[n] = params.channel_parameters[channel][var]
        
        # Write the id_timescale values
        temp_id_timescale = f.createVariable('id_timescale','i',('channels',))
        for (channel, n) in zip(channels, range(len(channels))):
            temp_id_timescale[n] = self.variables['id_timescale'][channel]
        
        # Laser pointing angle
        temp_v = f.createVariable('Laser_Pointing_Angle','d',('scan_angles',))
        temp_v[:] = params.general_parameters['Laser_Pointing_Angle']
        
        # Molecular calculation
        temp_v = f.createVariable('Molecular_Calc','i')
        temp_v[:] = params.general_parameters['Molecular_Calc']    

        # Laser pointing angles of profiles
        temp_v = f.createVariable('Laser_Pointing_Angle_of_Profiles','i',('time','nb_of_time_scales'))
        for (time_scale,n) in zip(self.variables['Raw_Data_Start_Time'],
                                  range(len(self.variables['Raw_Data_Start_Time']))):
            temp_v[:len(time_scale), n] = 0  # The lidar has only one laser pointing angle
        
        # Raw data start/stop time
        temp_raw_start = f.createVariable('Raw_Data_Start_Time','i',('time','nb_of_time_scales'))
        temp_raw_stop = f.createVariable('Raw_Data_Stop_Time','i',('time','nb_of_time_scales'))
        for (start_time, stop_time,n) in zip(self.variables['Raw_Data_Start_Time'],
                                             self.variables['Raw_Data_Stop_Time'],
                                             range(len(self.variables['Raw_Data_Start_Time']))):
            temp_raw_start[:len(start_time),n] = start_time
            temp_raw_stop[:len(stop_time),n] = stop_time

        #Laser shots
        temp_v =  f.createVariable('Laser_Shots','i',('time','channels'))
        for (channel,n) in zip(channels, range(len(channels))):
            time_length = len(self.variables['Raw_Data_Start_Time'][self.variables['id_timescale'][channel]])
            temp_v[:time_length, n] = params.channel_parameters[channel]['Laser_Shots']
        
        #Raw lidar data
        temp_v = f.createVariable('Raw_Lidar_Data','d',('time', 'channels','points'))
        for (channel,n) in zip(channels, range(len(channels))):
            c = self.channels[channel]
            temp_v[:len(c.time),n, :c.points] = c.matrix
        
        self.add_dark_measurements_to_netcdf(f, channels)
            
        #Pressure at lidar station
        temp_v = f.createVariable('Pressure_at_Lidar_Station','d')
        temp_v[:] = self.info['Pressure']

        #Temperature at lidar station
        temp_v = f.createVariable('Temperature_at_Lidar_Station','d')
        temp_v[:] = self.info['Temperature']   
        
        self.save_netcdf_extra(f)
        f.close()
   
    def add_dark_measurements_to_netcdf(self, f, channels):

        # Get dark measurements. If it is not given in self.dark_measurement
        # try to get it using the get_dark_measurements method. If none is found
        # return without adding something.
        if self.dark_measurement is None:
            self.dark_measurement = self.get_dark_measurements()
        
        if self.dark_measurement is None:
            return
        
        dark_measurement = self.dark_measurement
        
        # Calculate the length of the time_bck dimensions
        number_of_profiles = [len(c.time) for c in dark_measurement.channels.values()]
        max_number_of_profiles = np.max(number_of_profiles)
        
        # Create the dimension
        f.createDimension('time_bck', max_number_of_profiles)
         
        # Save the dark measurement data
        temp_v = f.createVariable('Background_Profile','d',('time_bck', 'channels', 'points'))
        for (channel,n) in zip(channels, range(len(channels))):
            c = dark_measurement.channels[channel]
            temp_v[:len(c.time),n, :c.points] = c.matrix
        
        # Dark profile start/stop time
        temp_raw_start = f.createVariable('Raw_Bck_Start_Time','i',('time_bck','nb_of_time_scales'))
        temp_raw_stop = f.createVariable('Raw_Bck_Stop_Time','i',('time_bck','nb_of_time_scales'))
        for (start_time, stop_time,n) in zip(dark_measurement.variables['Raw_Data_Start_Time'],
                                             dark_measurement.variables['Raw_Data_Stop_Time'],
                                             range(len(dark_measurement.variables['Raw_Data_Start_Time']))):
            temp_raw_start[:len(start_time),n] = start_time
            temp_raw_stop[:len(stop_time),n] = stop_time
        
        # Dark measurement start/stop time
        f.RawBck_Start_Date = dark_measurement.info['start_time'].strftime('%Y%m%d')
        f.RawBck_Start_Time_UT = dark_measurement.info['start_time'].strftime('%H%M%S')
        f.RawBck_Stop_Time_UT = dark_measurement.info['stop_time'].strftime('%H%M%S')



    def save_netcdf_extra(self, f):
        pass
        
    def _gettime(self, date_str, time_str):        
        t = datetime.datetime.strptime(date_str+time_str,'%d/%m/%Y%H.%M.%S')
        return t
    
    def plot(self):
        for channel in self.channels:
            self.channels[channel].plot(show_plot = False)
        plt.show()
    
    def get_dark_measurements(self):
        return None


class Lidar_channel:

    def __init__(self,channel_parameters):
        c = 299792458 #Speed of light
        self.wavelength = channel_parameters['name']
        self.name = str(self.wavelength)
        self.binwidth = float(channel_parameters['binwidth']) # in microseconds
        self.data = {}
        self.resolution = self.binwidth * c / 2
        self.z = np.arange(len(channel_parameters['data'])) * self.resolution + self.resolution/2.0 # Change: add half bin in the z 
        self.points = len(channel_parameters['data'])
        self.rc = []
        self.duration  = 60
        
    def calculate_rc(self):
        background = np.mean(self.matrix[:,4000:], axis = 1) #Calculate the background from 30000m and above
        self.rc = (self.matrix.transpose()- background).transpose() * (self.z **2)
        
    
    def update(self):
        self.start_time = min(self.data.keys())
        self.stop_time = max(self.data.keys()) + datetime.timedelta(seconds = self.duration)
        self.time = tuple(sorted(self.data.keys()))
        sorted_data = sorted(self.data.iteritems(), key=itemgetter(0))
        self.matrix =  np.array(map(itemgetter(1),sorted_data))
    
    def _nearest_dt(self,dtime):
        margin = datetime.timedelta(seconds = 300)        
        if ((dtime + margin) < self.start_time)| ((dtime - margin)   > self.stop_time):
            print "Requested date not covered in this file"
            raise
        dt = abs(self.time - np.array(dtime))
        dtmin = min(dt)

        if dtmin > datetime.timedelta(seconds = 60):
            print "Nearest profile more than 60 seconds away. dt = %s." % dtmin
        ind_t = np.where(dt == dtmin)
        ind_a= ind_t[0]
        if len(ind_a) > 1:
            ind_a = ind_a[0]
        chosen_time = self.time[ind_a]
        return chosen_time, ind_a
    
    def subset_by_time(self, start_time, stop_time):
        
        time_array = np.array(self.time)
        condition = (time_array >= start_time) & (time_array <= stop_time)
        
        subset_time = time_array[condition]
        subset_data = dict([(c_time, self.data[c_time]) for c_time in subset_time])
        
        #Create a list with the values needed by channel's __init__()
        parameters_values = {'name': self.wavelength, 
                             'binwidth': self.binwidth, 
                             'data': subset_data[subset_time[0]],}

        c = Lidar_channel(parameters_values)
        c.data = subset_data
        c.update()
        return c
        
    
    def profile(self,dtime, signal_type = 'rc'):
        t, idx = self._nearest_dt(dtime)
        if signal_type == 'rc':
            data = self.rc
        else:
            data = self.matrix
        
        prof = data[idx,:][0]
        return prof, t

    def get_slice(self, starttime, endtime, signal_type = 'rc'):
        if signal_type == 'rc':
            data = self.rc
        else:
            data = self.matrix
        tim = np.array(self.time)
        starttime = self._nearest_dt(starttime)[0]
        endtime = self._nearest_dt(endtime)[0]
        condition = (tim >= starttime) & (tim <= endtime)
        sl  = data[condition, :]
        t = tim[condition]
        return sl,t
    
    def profile_for_duration(self, tim, duration = datetime.timedelta(seconds = 0), signal_type = 'rc'):
        """ Calculates the profile around a specific time (tim). """
        starttime = tim - duration/2
        endtime = tim + duration/2
        d,t = self.get_slice(starttime, endtime, signal_type = signal_type)
        prof = np.mean(d, axis = 0)
        tmin = min(t)
        tmax = max(t)
        tav = tmin + (tmax-tmin)/2
        return prof,(tav, tmin,tmax)
    
    def average_profile(self):
        """ Return the average profile (NOT range corrected) for all the duration of the measurement. """
        prof = np.mean(self.matrix, axis = 0)
        return prof
    
    def plot(self, signal_type = 'rc', filename = None, zoom = [0,12000,0,-1], show_plot = True, cmap = plt.cm.jet, z0 = None, title = None, vmin = 0, vmax = 1.3 * 10 ** 7):
        #if filename is not None:
        #    matplotlib.use('Agg')
        
        fig = plt.figure()
        ax1 = fig.add_subplot(111)
        self.draw_plot(ax1, cmap = cmap, signal_type = signal_type, zoom = zoom, z0 = z0, vmin = vmin, vmax = vmax)
        
        if title:
            ax1.set_title(title)
        else:
            ax1.set_title("%s signal - %s" % (signal_type.upper(), self.name)) 
        
        if filename is not None:
            pass
            #plt.savefig(filename)
        else:
            if show_plot:
                plt.show()
        #plt.close() ???
    
    def draw_plot(self,ax1, cmap = plt.cm.jet, signal_type = 'rc', zoom = [0,12000,0,-1], z0 = None, cmap_label = 'a.u.', vmin = 0, vmax = 1.3 * 10 ** 7):
        
        if signal_type == 'rc':
            if len(self.rc) == 0:
                self.calculate_rc()
            data = self.rc
        else:
            data = self.matrix
        
        hmax_idx = self.index_at_height(zoom[1])
        
        # If z0 is given, then the plot is a.s.l.
        if z0:
            ax1.set_ylabel('Altitude a.s.l. [km]')
        else:
            ax1.set_ylabel('Altitude a.g.l. [km]')
            z0 = 0
            
        ax1.set_xlabel('Time UTC [hh:mm]')
        #y axis in km, xaxis /2 to make 30s measurements in minutes. Only for 1064
        #dateFormatter = mpl.dates.DateFormatter('%H.%M')
        #hourlocator = mpl.dates.HourLocator()
        
        #dayFormatter = mpl.dates.DateFormatter('\n\n%d/%m')
        #daylocator = mpl.dates.DayLocator()       
        hourFormatter = mpl.dates.DateFormatter('%H:%M')
        hourlocator = mpl.dates.AutoDateLocator(interval_multiples=True)

        
        #ax1.axes.xaxis.set_major_formatter(dayFormatter)
        #ax1.axes.xaxis.set_major_locator(daylocator)
        ax1.axes.xaxis.set_major_formatter(hourFormatter)
        ax1.axes.xaxis.set_major_locator(hourlocator)

        
        ts1 = mpl.dates.date2num(self.start_time)
        ts2 = mpl.dates.date2num(self.stop_time)
        
        
        im1 = ax1.imshow(data.transpose()[zoom[0]:hmax_idx,zoom[2]:zoom[3]],
            aspect = 'auto',
            origin = 'lower',
            cmap = cmap,
            vmin = vmin,
            #vmin = data[:,10:400].max() * 0.1,
            vmax = vmax,
            #vmax = data[:,10:400].max() * 0.9,
            extent = [ts1,ts2,self.z[zoom[0]]/1000.0 + z0/1000., self.z[hmax_idx]/1000.0 + z0/1000.],
            )
        
        cb1 = plt.colorbar(im1)
        cb1.ax.set_ylabel(cmap_label)

    def index_at_height(self, height):
        idx = np.array(np.abs(self.z - height).argmin())
        if idx.size >1:
            idx =idx[0]
        return idx
    
def netcdf_from_files(LidarClass, filename, files, channels, measurement_ID):
    #Read the lidar files and select channels
    temp_m = LidarClass(files)
    m = temp_m.subset_by_channels(channels)
    m.get_PT()
    m.info['Measurement_ID'] = measurement_ID
    m.save_as_netcdf(filename)