HyTraj

This python library implements HySPLIT based trajectory modeling and analysis.

Installation

Install using pip:

pip install hytraj

or

pip install git+https://github.com/pankajkarman/HyTraj.git

Dependencies

1. Plotting requires Basemap.

2. Hierarchical clustering requires traj_dist.

Documentation

Latest documentation is available here.

Usage

See this notebook for example usecase.

import hytraj as ht

Generate Trajectories

from hytraj import HyTraj

met_type = "ncep"
dates = pd.date_range("2010-02-01", freq="24H", end="2010-02-10")
hy = HyTraj(stations, height, run_time, working, metdir, outdir, met_type)
data = hy.run(dates, njobs=7)
hy.plot(data["Neumayer"], vertical="alt", show=True)

Cluster Trajectories

KMeans Clustering using wavelet features

from hytraj import HyCluster

labels = HyCluster(data).fit(kmax=10, method='KMeans')

Hierarchical Agglomerative Clustering (HAC)

from hytraj import HyHAC

trj = HyHAC(data)
labels = trj.fit(nclus=4, metric='sspd')
trj.plot_dendrogram()

Receptor Modeling

from hytraj import HyReceptor, HyData

station = 'South Pole'
data = HyData(files, stations).read()[station]
model = HyReceptor(ozone, data, station_name="South Pole")
cwt = model.calculate_cwt(weighted=False)
pscf = model.calculate_pscf(thresh=0.95)
rtwc = model.calculate_rtwc(normalise=True)
model.plot_map(rtwc, boundinglat=-25)

Features

1. HyTraj: Higher level implementation of Parallel Generation, reading and plotting of Trajectories (Recommended).

2. HyGen: Generation of Trajectories using various meteo datasets (NCEP and GDAS implemented).

3. HyControl: Generation of control files for parallel trajectory generation afterwards.

4. HyParallel: Parallel generation of trajectories using control files produced using HyControl.

5. HyData: Reading and binning trajectories data (NetCDF with xarray support).

6. HyCluster: Clustering of trajectories with KMeans using wavelet features.

7. HyHAC: Clustering of trajectories with Hierarchical Agglomerative Clustering (HAC) using various trajectory distance metric like DTW, EDR, LCSS, SSPD, Frechet Distance, Hausdorf Distance.

8. HyReceptor: Single site Receptor Modeling ( both weighted and unweighted):

   • Concentration weighted Trajectory (CWT)
   • Potential Source Contribution Function (PSCF)
   • Residence Time Weighted Concentration (RTWC)

To Do

1. Support for more meteorology like ERA5.
2. Add documentation.
3. GUI: Medium-term goal

PS: Find pre-built HYSPLIT executable at this link and copy executeble to working directory.