def grab_prob(time_idx):
    # make frame__idx an integer to avoid slicing errors
    frame_idx = int(time_idx)
    # get 'frame_time'
    frame_time = grab_sst_time(frame_idx)

    # set month of year
    MoY_val = int(frame_time.month)

    # get list of temperature and salinity values from subset
    temp = []
    salt = []

    point_list = zip(eta_rho,xi_rho

    # append values
    for i, j in point_list:
    	temp.append(fh.variables['temp'][frame_idx,29,i,j])
    	salt.append(fh.variables['salt'][frame_idx,29,i,j])

    data = {'var1': temp, 'var2': salt}
    data = pd.DataFrame(data=data)
    data['MoY'] = MoY_val
    # remove masked floats
    data = data[data.var1 >= 1]

    # calculate probabilities
    probs = lr_model.predict_proba(data[['var1','var2','MoY']])
    prob_TSW, prob_EAC = zip(*probs)
    # convert tuples to list
    prob_EAC = list(prob_EAC)
    # make 1D array
    prob_EAC = np.asarray(prob_EAC)

    # calulcate ratio metric
    count_EAC = np.count_nonzero(prob_EAC > 0.5)
    count_TSW = np.count_nonzero(prob_EAC < 0.5)

    return count_EAC, xtime # << return these instead of modifying a global var



 ## ....
        for i in range(0, len(time)):
          count_EAC, xtime = grab_prob(i)
          ## add them to the data frame