What does the config.yml contain?

Instead of importing the params as an executable module as was previously the case, the workflow now relies on importing a YAML file and running that through a function to parse the inputs into usable objects by downstream functions. This is mostly to address potential security concerns and to facilitate runs that incorporate multiple geometries. The YAML config file is read as a dictionary with values being checked and/or updated by the parse_params function in the run_ltop_complete.py script. This dictionary contains all of the necessary components to run the full workflow. There are a few optional arguments that are currently (10/18/2022) set as defaults in various functions. See below for notes on those.

run_times: str: One of either ‘multiple’ or ‘single’. Multiple will set it up to run multiple ROIs in the same call while single assumes that you are going to just pass one ROI and run the workflow once. The script will error if you pass a different term than single or multiple.
place: str: This will be used in naming construction throughout the process so give it a short, meaningful name. Note that if you are going to run multiple geometries in a for loop this will be overwritten by the process that starts those runs.
assetsRoot: str: This will be either your ‘users/username’ or moving to the future requirements for GEE it should be linked to a gcloud project and would look like: ‘projects/name-of-your-project/assets/’, for example.
assetsChild: str: The folder you want your assets to end up in inside your project. Note that the script will create this folder if it does not find it in your project repo.
version:: 0.1.1 as of 10/18/2022 with the 0.1 designating the shift to Python.
startYear: int: Most of the scripts assume 1990 is the start year. You can run it with a different time frame but generally we advise running as many years as possible in the LandTrendr algorithm.
endYear: int: Most recent full year available in input imagery
seedSpacing: int: This defines the starting seed spacing for the SNIC algorithm. The default is 10. See GEE documentation for more information on this param. It only needs to be changed if you are interested in the specific effects of the SNIC inputs/outputs.
randomPts: int: The number of random points we use to sample the SNIC imagery. There is not a best known value but generally more is better. Default is 20,000. You can try more, we don’t necessarily recommend less but this depends on the area you are running. This seemed sufficient for an area the size of SE Asia.
imageSource: str: One of either ‘servir’ or ‘medoid’. The default is that anything not specified as medoid will choose servir composites. Note that as of 10/18/2022 the functionality for medoids does not currently exist.
assetsRoot: str: Defined from assetsRoot variable above.
assetsChild: str: Defined from assetsChild above.
aoi: GEE FC: This is the study area you want to run. NOTE that it must be cast as a .geometry() object, not just a featureCollection.
maxClusters: int: Defines the max clusters for the kmeans algorithm implemented in GEE. Somewhere around 5,000 seems to be the max. Generally, 2500 seems sufficient for something about the size of Cambodia.
minClusters: int: Generally set to the same as maxClusters but could be set lower.
selectedLTparams: GEE FC: Leave defaults and this will just be constructed for you from the other input params.
param_scoring_inputs: str, filepath like: Filepath to a local directory where you want the LT outputs from running on abstract image points to reside. Note that this folder will be created for you in the script if you have not created it in advance. This will be dynamically updated in the code if you chose ‘multiple’ for the run_times arg above.
outfile: str, filepath like: This should be a local directory where you want the selected LT params for generating LTOP breakpoints to end up. Note that the folder will be created for you if you have not created it in advance. This will be dynamically updated in the code if you chose ‘multiple’ for the run_times arg above.
njobs: int: Number of cores to use in param selection. This will likely be changed in future versions.
cloud_bucket: str, GCS cloud bucket name: This is the name of a cloud bucket in a Google Cloud Services account. See notes in general setup <general_setup.rst> for setting this up. Note that the program will not do this for you, you need to manually do this in advance of running or you will hit an error.

NOTE that as of 10/18/2022, the necessary code to run medoid composites does not yet exist. When that is done, the below args will be required to generate medoid composites from existing LandTrendr modules. startDate: str A string date range formatted like ‘11-20’ (mm-dd). endDate: str A string date range formatted like ‘03-10’ (mm-dd). masked: list A list of things you want to mask formatted like [‘cloud’, ‘shadow’], for example.

Other arguments

The class that runs LTOP takes the param dictionary as the primary input, passed as *args, and then a sleep_time argument which can be passed separately to the class. This defines the time you want the program to wait before re-checking if a process or task is complete. There is not a ‘perfect’ answer here but it is mostly in place so that the program does not try to query Google’s servers every iteration of a while loop.