Skip to main content

Generates time-varying weather profiles using a synthesis of in-situ observations and satellite reanalysis estimates of meteorological parameters.

Project description


weathergen generates time-varying weather profiles using a synthesis of in-situ observations and satellite reanalysis estimates of meteorological parameters, and is able to recreate weather patterns allowing for the simulation frameworks like maria to simulate climatologically-accurate ground-based astronomical observations.


Atmospheric parameters are generated by combining three ingredients:

  1. The probability distributions of each parameter for each time of year and time of day.

  2. An eigenmodal decomposition of how parameters covary.

  3. Characteristic spectra of how each eigenmode changes over time.

These are all calculated using climatological data from ERA5, GML, ESO, MKWC, and Meteostat. For each region, reanalysis weather parameters are compiled and adjusted (e.g. for seasonal diurnal and trends) with respect to in-situ meteorological trends.

The generated profiles depend on reanalysis data which are limited to a temporal resolution of an hour, but other means of adjustment are possible depending on the sophistication of the alternate sources of weather data. For example, the level of turbulent fluctuation in total column water vapor (the largest driver of sub-millimeter atmospheric interference) is not stationary from hour to hour.


Install the package using pip:

pip install weathergen

Import the package and instantiate a `Weather` object, and pass it an array of times to the `generate` function. For example, to simulate hourly weather for Princeton, New Jersey between June 1st and September 1st we would write

import numpy as np
from datetime import datetime
import weathergen

t0 = datetime(2022,6,1).timestamp() # June 1, 2022
t1 = datetime(2022,9,1).timestamp() # August 1, 2022

gen_times = np.arange(t0, t1, 3600) # three months of hourly data

weather = weathergen.Weather(region='princeton', seasonal=True, diurnal=True)
weather.generate(time=gen_times, mode='random')

Specified times should be supplied in Unix time. The weather parameters are contained in the attributes of the weather object, e.g. weather.temperature or weather.pressure. The values are typically two-dimensional with shape (n_height, n_time), where weather.height and weather.time describe the time and height of each dimension, in Unix time and meters above sea level. Single-level parameters are described by a single number for each time and do not have a height dimension.

Note that the supplied year is arbitrary; the underlying model considers only seasonal and diurnal climatological variations. All available regions are outlined in the dataframe weathergen.regions.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

weathergen-2.4.0.tar.gz (91.7 MB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page