CC_Est_NARC/Analysis/Code/climdata_fcts.py

# -*- coding: utf-8 -*-
"""
Created on Thu Jun 14 16:32:01 2018

@author: z5025317
"""

import matplotlib.pyplot as plt
from datetime import datetime, timedelta
import numpy as np
import pandas as pd



def compare_images(im1, im2):
    """plots 2 images next to each other, sharing the axis"""
    plt.figure()
    ax1 = plt.subplot(121)
    plt.imshow(im1, cmap='gray')
    ax2 = plt.subplot(122, sharex=ax1, sharey=ax1)
    plt.imshow(im2, cmap='gray')
    plt.show()  
    
def reject_outliers(data, m=2):
    "rejects outliers in a numpy array"
    return data[abs(data - np.mean(data)) < m * np.std(data)]

def duplicates_dict(lst):
    "return duplicates and indices"
    # nested function
    def duplicates(lst, item):
            return [i for i, x in enumerate(lst) if x == item]
        
    return dict((x, duplicates(lst, x)) for x in set(lst) if lst.count(x) > 1)

def datenum2datetime(datenum):
    "convert datenum to datetime"
    #takes in datenum and outputs python datetime
    time = [datetime.fromordinal(int(dn)) + timedelta(days=float(dn)%1) - timedelta(days = 366) for dn in datenum]
    return time




def select_min_med_max_dif_model(NARCLIM_df):
    #Select the 3 most representative models (min med and max difference betwen far future and present)
    Fdf_1900_2080_sorted = NARCLIM_df.reindex_axis(sorted(NARCLIM_df.columns), axis=1)
    Fdf_1900_2080_sorted_means = pd.DataFrame(Fdf_1900_2080_sorted.mean())
    df = Fdf_1900_2080_sorted_means
    #add a simple increasing integer index 
    df = df.reset_index()
    df= df[df.index % 3 != 1]
    df['C'] = df[0].diff()
    df = df.reset_index()
    df= df[df.index % 2 != 0]
    #get max difference model (difference between far future and prsent day)
    a = df[df.index == df['C'].argmax(skipna=True)]
    Max_dif_mod_name = a.iloc[0]['index']
    #get min difference model
    a = df[df.index == df['C'].argmin(skipna=True)]
    Min_dif_mod_name = a.iloc[0]['index']
    #get the model which difference is closest to the median difference
    df['D'] = abs(df['C']- df['C'].median())
    a = df[df.index == df['D'].argmin(skipna=True)]
    Med_dif_mod_name = a.iloc[0]['index']
    #data frame with min med and max difference model
    df2 = NARCLIM_df.filter(regex= Min_dif_mod_name[:-5] + '|' +  Med_dif_mod_name[:-5] + '|' +  Max_dif_mod_name[:-5] )
    dfall = df2.reindex_axis(sorted(df2.columns), axis=1)
    #data frame with individual models
    dfmin = NARCLIM_df.filter(regex= Min_dif_mod_name[:-5])
    dfmax = NARCLIM_df.filter(regex= Max_dif_mod_name[:-5])
    dfmed = NARCLIM_df.filter(regex= Max_dif_mod_name[:-5])
    return dfall , dfmin, dfmed, dfmax, Min_dif_mod_name,Med_dif_mod_name, Max_dif_mod_name


def calculate_deltas_NF_FF2(Annual_df, Seasonal_df):
    """calculates the "deltas" between nearfuture and present day for annual or seasonal climate data in pandas TS format"""

    times = ['annual', 'DJF', 'MAM', 'JJA','SON']
    delta_all_df = pd.DataFrame()
    for temp in times:
        if temp == 'annual':
            Mean_df = Annual_df.mean()
            Column_names = ['near', 'far']
        if temp == 'DJF':
            Mean_df = Seasonal_df[Seasonal_df.index.quarter==1].mean()
            Column_names = ['DJF_near', 'DJF_far']
        if temp == 'MAM':
            Mean_df = Seasonal_df[Seasonal_df.index.quarter==2].mean()
            Column_names = ['MAM_near', 'MAM_far']
        if temp == 'JJA':
            Mean_df = Seasonal_df[Seasonal_df.index.quarter==3].mean()
            Column_names = ['JJA_near', 'JJA_far']
        if temp == 'SON':
            Mean_df = Seasonal_df[Seasonal_df.index.quarter==4].mean()
            Column_names = ['SON_near', 'SON_far']
        models = list(Seasonal_df.mean().index)
        newmodel = []
        type(newmodel)
        for each in models:
            newmodel.append(each[:-5])
        unique_models = set(newmodel)
        # calculate diff for each unique model
        delta_NF_ensemble = []
        delta_FF_ensemble = []
        for unique_model in unique_models:
            dfdiff = Mean_df.filter(regex= unique_model)
            type(dfdiff)
            delta_NF = dfdiff[1] - dfdiff[0]
            delta_NF_ensemble.append(delta_NF)
            delta_FF = dfdiff[2] - dfdiff[1]
            delta_FF_ensemble.append(delta_FF)
    
        delta_df1=pd.DataFrame(delta_NF_ensemble, index=unique_models)
        delta_df2=pd.DataFrame(delta_FF_ensemble, index=unique_models)
        delta_df= pd.concat([delta_df1, delta_df2], axis=1)
    
        #rename columns
        delta_df.columns = Column_names
        #add a row with medians and 10 and 90th percentiles
        delta_df.loc['10th'] = pd.Series({Column_names[0]:np.percentile(delta_df[Column_names[0]], 10), Column_names[1]:np.percentile(delta_df[Column_names[1]], 10)})
        delta_df.loc['median'] = pd.Series({Column_names[0]:np.percentile(delta_df[Column_names[0]], 50), Column_names[1]:np.percentile(delta_df[Column_names[1]], 50)})
        delta_df.loc['90th'] = pd.Series({Column_names[0]:np.percentile(delta_df[Column_names[0]], 90), Column_names[1]:np.percentile(delta_df[Column_names[1]], 90)})
        #append df to overall df
        delta_all_df = pd.concat([delta_all_df, delta_df], axis=1)
        return delta_all_df