Skip to contents

duration

Production code for CMS Duration Estimation

CI lint CI tests Check Link Rot GitHub license GitHub contributors

Estimating Methane Emission Durations

This repository contains code used to estimate methane emission durations using concentration observations from a network of point-in-space continuous monitoring systems (CMS). The scripts include:

  • estimate_durations.R: main code for estimating durations
  • helpers.R: suite of helper functions sourced internally in estimate_durations()

The accompanying paper can be found here: https://doi.org/10.1021/acs.estlett.4c00687

Note: The code includes a shortened “toy” set of data from the 2023 ADED controlled release experiment. This is meant for demonstration purposes only to get users started with the code and to learn the required structure of input data.

Installation & Usage

Access the latest (dev) version:

#install.packages("devtools") # if needed on first usage

devtools::install_github("Hammerling-Research-Group/duration")

library(duration)

Once installed and loaded, users should start with the core function, which produces duration estimates by emission source.

For example, run the following line to generate emission duration estimates, return a plot of the distributions of duration estimates by source using the supplied “toy” data, and store results on your Desktop (note: full output is saved by default, the location of which is shared with the user):

estimate_durations(plot = TRUE, directory = "Desktop")

Some Notes to Consider

The default values are meant to allow for simple usage without the need for adjustment. Yet, given that toy data are used as input by default, at a minimum users should adjust the data by supplying real site data.

Available arguments for estimate_durations() include:

  • data: Forward model output data stored as a list object. The format must match that of the sample input data in the duration package, or output data from other similar implementations, e.g., from MAIN_1_simulate.R (more details on data structure are in the following section)
  • tz: Time zone. Default set to America/Denver
  • directory: Location where results are saved. If not specified, a directory is created and the location is shared with the user
  • events: Print the number of events per year for a quick check? Default set to FALSE
  • plot: Return a plot of the distribution of average duration estimates by source? Default set to FALSE
  • time: Return the timing of the simulation? Default set to TRUE

Input Data Structure

A key piece of working reliably with the duration package is the structure of the input data, which is very specific. To get a sense of what is required, users can read in and explore the provided sample data to learn about the structure, which must be contained in a list object. Each of the following basic options should be sufficient to allow users to structure their input data accordingly in order for the code to work as intended, and for the results to be reliable.

  1. Read in the data:
df <- list.files(system.file("input_data", package = "duration"), full.names = TRUE)
data <- readr::read_rds(df[1])
  1. Once read in, users can explore the data directly:
View(data)
  1. Check out which objects are stored in the list for reference:
> names(data)

[1] "times"   "obs"   "WD"    "WS"    "Tank"    "Separator.East"    "Wellhead.East"   "Wellhead.West"   "Separator.West"
  1. Or, inspect the structure of the input data directly:
> str(data)

List of 9
 $ times         : POSIXct[1:4001], format: "2023-02-21 21:19:00" "2023-02-21 21:20:00" "2023-02-21 21:21:00" "2023-02-21 21:22:00" ...
 $ obs           :'data.frame': 4001 obs. of  10 variables:
  ..$ ENE: num [1:4001] 2.02 2.04 2.03 2.06 2.01 ...
  ..$ ESE: num [1:4001] 2.25 2.25 NA 2.24 2.24 ...
 $ WD            : num [1:4001] 1.98 1.98 2.02 2.04 1.93 ...
 $ WS            : num [1:4001] 3.41 3.52 3.25 2.71 2.72 2.94 3.08 2.81 3.16 3.07 ...
 $ Tank          :'data.frame': 4001 obs. of  10 variables:
  ..$ ENE: num [1:4001] 1.86e-49 3.11e-49 2.44e-49 1.32e-50 2.62e-85 ...
  ..$ ESE: num [1:4001] 1.81e-59 2.50e-59 2.12e-59 1.61e-60 9.60e-104 ...
 $ Separator.East:'data.frame': 4001 obs. of  10 variables:
  ..$ ENE: num [1:4001] 6.41e-43 8.05e-43 6.14e-43 2.61e-43 1.04e-75 ...
  ..$ ESE: num [1:4001] 1.35e-55 1.42e-55 1.87e-55 1.77e-55 6.67e-98 ...
 $ Wellhead.East :'data.frame': 4001 obs. of  10 variables:
  ..$ ENE: num [1:4001] 4.00e-24 5.48e-24 1.71e-24 9.56e-26 1.83e-40 ...
  ..$ ESE: num [1:4001] 1.26e-37 1.66e-37 1.07e-37 3.09e-38 3.09e-66 ...
 $ Wellhead.West :'data.frame': 4001 obs. of  10 variables:
  ..$ ENE: num [1:4001] 7.44e-90 1.47e-89 8.16e-90 5.46e-91 1.21e-151 ...
  ..$ ESE: num [1:4001] 1.95e-104 3.68e-104 2.18e-104 3.86e-105 1.78e-177 ...
 $ Separator.West:'data.frame': 4001 obs. of  10 variables:
  ..$ ENE: num [1:4001] 2.79e-60 1.03e-59 9.98e-60 5.54e-62 2.39e-101 ...
  ..$ ESE: num [1:4001] 6.40e-69 2.46e-68 2.46e-68 1.97e-70 5.61e-117 ...
  1. Specific information about each component of the input list is given below:
  • times: Times of the CMS concentration data. Times must be at regular intervals, ideally every minute. Time information must be saved in a string format that can be read by the as_datetime function of the lubridate package, e.g., "yyyy-mm-ddThh:mm:ss".
  • obs: CMS concentration observations. The rows of the data frame correspond to the times of the observations and the columns correspond to the different CMS sensor locations.
  • WD: Wind direction measurements in radians. The length of this vector must match the length of times.
  • WS: Wind speed measurements in m/s. The length of this vector must match the length of times.
  • Tank: Data frame containing simulated concentrations at the CMS sensor locations assuming that the tank is the emission source. The rows of the data frame correspond to the times of the simulated observations and the columns correspond to the different CMS sensor locations. A variable like Tank is required for all potential emission sources on the site.