Example for cross sectional flux method¶

%matplotlib inline

import os
import numpy as np
import pandas as pd
import scipy.stats
import xarray as xr

import ddeq
import ucat

# Source dataset
sources = xr.Dataset({
    "source": xr.DataArray(["6298"], dims="source"),
    "label": xr.DataArray(["Janschwalde"], dims="source"),
    "lon": xr.DataArray([14.457959], dims="source"),
    "lat": xr.DataArray([51.836241], dims="source"),
    "diameter": xr.DataArray([1000.0], dims="source"),
    "NOx_emissions": xr.DataArray([0.42778918548939077], dims="source"), # Bottom-up estimate of emissions (in kg/s)
})

# Setup domain for plotting
NAME = "6298" # CORSO ID for Janschwalde power plant
SOURCE = sources.sel(source=[NAME])
DOMAIN = ddeq.misc.Domain.around_source(SOURCE)
CRS = ddeq.misc.get_opt_crs(DOMAIN)

# CSF settings going 30 km downstream and 80 km across
DMAX = 30e3
PLUME_WIDTH = 80e3

for chunk in ddeq.sats.iter_S5P_swath(filename, gas="NO2", preprocessed=False):

    # Plume detection using wind vector
    time = pd.Timestamp(chunk.attrs['time'])
    lvl_filename = time.strftime('/input/ERA5/CORSO/raw/ERA5-lvl-%Y%m%dt0000.nc')
    sng_filename = time.strftime('/input/ERA5/CORSO/raw/ERA5-sng-%Y%m%dt0000.nc')

    winds = ddeq.era5.read(
        sng_filename,
        lvl_filename,
        method='gnfra',
        times=time,
        extent=DOMAIN.extent,
        sources=SOURCE
    )

    chunk = ddeq.dplume.detect_from_wind(chunk, SOURCE, winds, dmax=DMAX, crs=CRS)

    if "x_nodes" in chunk: # center line was computed, i.e. Janschwalde is inside chunk
        break

# Prepare local coordinate system, plume area and convert from µmol/m² to kg/m²
data = ddeq.curves.compute_natural_coords(chunk)
data = ddeq.curves.compute_plume_areas(data, plume_width=PLUME_WIDTH)
ddeq.emissions.convert_units(data, "NO2", "NO2")

# Prepare emissions dataset (as code below assumes a time series)
emissions = xr.concat([emissions.copy()], dim="time")
N = emissions.time.size

# Line density ensemble
q = np.squeeze(emissions.NO2_line_density.values)
q_std = np.squeeze(emissions.NO2_line_density_precision.values)
q_sample = q + q_std * np.random.randn(10000)

# Wind speed ensemble using Log-normal distribution with mean of u_eff and RMSE of about 1.0 m/s
u = emissions.wind_speed.values.flatten()
u_std = 1.0
mu = np.log(u**2 / np.sqrt(u**2 + u_std**2))
sigma = np.sqrt(np.log(1 + u_std**2 / u**2))
u_sample = np.exp(mu + sigma * np.random.randn(10000, N))

# Compute NO2-to-NOx conversion (between 1 km and 30 km downstream of source
along = np.linspace(1, 30, 200) * 1e3

# NO2-to-NOx conversion ensemble
seconds_since_emission = along[None,:] / u[:,None]
seconds_sample = along[None,:,None] / u_sample[:,None,:]

NO2toNOx_model = ddeq.functions.NO2toNOxConversion(name="Janschwalde")

f_local, f_local_std = NO2toNOx_model(seconds_since_emission)
f_local_sample, _ = NO2toNOx_model(seconds_sample)
f_local_sample = f_local_sample.mean(1) + f_local_std.mean(1)[None,:] * np.random.randn(10000,N)

# NOx lifetime ensemble with median of 2.5 hours
tau = 2.5 * 60**2     # in seconds
tau_std = 0.8 * 60**2 # in seconds
mu = np.log(tau**2 / np.sqrt(tau**2 + tau_std**2))
sigma = np.sqrt(np.log(1 + tau_std**2 / tau**2))
tau_local_sample = np.exp(mu + sigma * np.random.randn(10000, N))

# Decay correction term
D_local_sample = np.exp(-along[None,:,None] / (tau_local_sample[:,None,:] * u_sample[:,None,:])).mean(1)

# NOx correction term
c_local_sample = f_local_sample / D_local_sample

# Create xarray dataset
sample = xr.Dataset()
sample["time"] = xr.DataArray(emissions.time.values, dims="time")
sample["model"] = xr.DataArray(["local"], dims="model")

sample["wind_speed"] = xr.DataArray(u_sample, dims=("n", "time"))

dims = ("model", "n", "time")
sample["NOx_NO2_ratio"] = xr.DataArray([f_local_sample], dims=dims, attrs={"models": ("Janschwalde",)})
sample["NOx_lifetime"] = xr.DataArray([tau_local_sample], dims=dims, attrs={"units": "s"})
sample["NOx_decay_correction"] = xr.DataArray([D_local_sample], dims=dims)
sample["NOx_correction"] = xr.DataArray([c_local_sample], dims=dims)