Tutorial: Batch Processing on Windows

Introduction

These are the steps that will need to be taken in order to use the batch scripting framework for InVEST models available in the natcap.invest python package.

For those wishing to do batch-processing with InVEST without setting up a Python scripting environment, see The InVEST CLI for examples of how to run InVEST models from the command-line.

Setting up your Python environment

We recommend using the conda or mamba package manager to create an environment. The easiest way to do so is to execute the following command:

mamba create -n invest-env -c conda-forge natcap.invest

This will create a new environment called invest-env with the latest released versions of python and natcap.invest, as well as all dependencies needed to execute natcap.invest.

Alternate installation methods are available. Please see the Installing the InVEST Python Package guide for further instructions on setting up your python environment and installing the natcap.invest package.

Creating Python Scripts

  1. Launch InVEST Workbench and Open the Selected Model

    Once you have selected an InVEST model to script, launch the workbench and select the model from the list of available models. This guide will follow the Nutrient Delivery Ratio (NDR) model.

  2. Fill in InVEST Model Input Parameters

    Once the user interface loads, populate the inputs in the model likely to be used in the Python script. For testing purposes the default InVEST’s data is appropriate. However, if you wish to write a script to execute several InVEST model runs in sequence, it would be reasonable to populate the user interface with data for the first run in the sequence.

  3. Generate a Python Script from the User Interface

    In the model interface, select “Save as…” in the left-hand menu bar, “Python script” in the dialog box, and then save the file to a known location.

    workbench_save_as.png modal_save_as.png

  1. Execute the script in the InVEST Python Environment

    Launch a shell of your choice and activate your InVEST environment you created earlier. We will use the Windows command prompt cmd.exe for an example, but the same commands should work on Powershell, bash and zsh. If you created an environment called invest-env, you would activate it with conda like so:

    > conda activate invest-env
    

    You will know your environment activated correctly when you see the environment name at the beginning of your shell prompt. For example:

    (invest-env) C:\Users\demo >
    

    Once the environment has been activated, we can run the python script we saved in the previous step. Suppose we’re on Windows and the script was saved to C:\Users\demo\Desktop\ndr.py, the command to run the script would be:

    (invest-env) C:\Users\demo > python C:\Users\demo\Desktop\ndr.py
    
  2. Output Results

    As the model executes, status information will be printed to the console. Once complete, model results can be found in the workspace folder selected during the initial configuration.

Modifying a Python Script

InVEST Python scripts consist of two sections:

  • The argument dictionary that represents the model’s user interface input boxes and parameters.

  • The call to the InVEST model itself.

For reference, consider the following script generated by the Nutrient model with default data inputs:

# coding=UTF-8
# -----------------------------------------------
# Generated by InVEST 3.14.2 on Wed Oct 23 10:49:40 2024
# Model: Nutrient Delivery Ratio

import logging
import sys

import natcap.invest.ndr.ndr
import natcap.invest.utils

LOGGER = logging.getLogger(__name__)
root_logger = logging.getLogger()

handler = logging.StreamHandler(sys.stdout)
formatter = logging.Formatter(
    fmt=natcap.invest.utils.LOG_FMT,
    datefmt='%m/%d/%Y %H:%M:%S ')
handler.setFormatter(formatter)
logging.basicConfig(level=logging.INFO, handlers=[handler])

args = {
    'biophysical_table_path': 'C:\\Users\demo\invest-sample-data\\NDR\\biophysical_table_gura.csv',
    'calc_n': True,
    'calc_p': True,
    'dem_path': 'C:\\Users\demo\invest-sample-data\\NDR\\DEM_gura.tif',
    'k_param': '2',
    'lulc_path': 'C:\\Users\demo\invest-sample-data\\NDR\land_use_gura.tif',
    'results_suffix': 'gura',
    'runoff_proxy_path': 'C:\\Users\demo\invest-sample-data\\NDR\precipitation_gura.tif',
    'subsurface_critical_length_n': '200',
    'subsurface_eff_n': '0.8',
    'threshold_flow_accumulation': '1000',
    'watersheds_path': 'C:\\Users\demo\invest-sample-data\\NDR\watershed_gura.shp',
    'workspace_dir': 'C:\\Users\demo\invest-sample-data\\NDR\workspace',
}

if __name__ == '__main__':
    natcap.invest.ndr.ndr.execute(args)

Elements to note:

  • Parameter Python Dictionary: Key elements include the 'args' dictionary. Note the similarities between the key values such as 'workspace_dir' and the equivalent “Workspace” input parameter in the user interface. Every key in the 'args' dictionary has a corresponding reference in the user interface.

    In the example below we’ll modify the script to execute the nutrient model for a parameter study of 'threshold_flow_accumulation'.

  • Execution of the InVEST model: The InVEST API invokes models with a consistent syntax where the module name that contains the InVEST model is listed first and is followed by a function called ‘execute’ that takes a single parameter called 'args'. This parameter is the dictionary of input parameters discussed above. In this example, the line

    natcap.invest.ndr.ndr.execute(args)
    

    executes the nutrient model end-to-end. If the user wishes to make batch calls to InVEST, this line will likely be placed inside a loop.

Example: Threshold Flow Accumulation Parameter Study

This example executes the InVEST NDR model on 10 values of threshold accumulation stepping from 500 to 1000 pixels in steps of 50. To modify the script above, replace the execution call with the following loop:

if __name__ == '__main__':
    #Loops through the values 500, 550, 600, ... 1000
    for threshold_flow_accumulation in range(500, 1001, 50):
        #set the accumulation threshold to the current value in the loop
        args['threshold_flow_accumulation'] = threshold_flow_accumulation
        #set the suffix to be accum### for the current threshold_flow_accumulation
        args['suffix'] = 'accum' + str(threshold_flow_accumulation)
        natcap.invest.ndr.ndr.execute(args)

This loop executes the InVEST nutrient model 10 times for accumulation values 500, 550, 600, ... 1000 and adds a suffix to the output files so results can be distinguished.

Example: Invoke NDR Model on a directory of Land Cover Maps

In this case we invoke the InVEST nutrient model on a directory of land cover data located at C:UserRichDesktoplandcover_data. As in the previous example, replace the last line in the UI generated Python script with:

import os
landcover_dir = r'C:\User\demo\Desktop\landcover_data'

if __name__ == '__main__':
    #Loop over all the filenames in the landcover dir
    for landcover_file in os.listdir(landcover_dir):
        #Point the landuse uri parameter at the directory+filename
        args['lulc_uri'] = os.path.join(landcover_dir, landcover_file)
        #Make a useful suffix so we can differentiate the results
        args['suffix'] = 'landmap' + os.path.splitext(landcover_file)[0]
        #call the nutrient model
        natcap.invest.ndr.ndr.execute(args)

This loop covers all the files located in C:\User\demo\Desktop\landcover_data and updates the relevant lulc_uri key in the args dictionary to each of those files during execution as well as making a useful suffix so output files can be distinguished from each other.

Example: Saving model log messages to a file

When you save a model run to a python script, InVEST will automatically include code to write all logging to your console while it executes. But sometimes, it’s useful to save log messages from the model to a file. This can be done by adding a few lines of code to the script, just after the existing logging code:

file_handler = logging.FileHandler('logfile.txt')
root_logger.addHandler(file_handler)

This will capture all logging generated by any python package that executes while running the ndr model writing all messages to logfile.txt. While this is a common use case, the logging package provides functionality for many more complex logging features. For more advanced use of the python logging module, refer to the Python project’s Logging Cookbook

Example: Enabling Parallel Processing

Note

This is an in-development feature and should be used with caution.

Most InVEST models accept an optional entry in the args dictionary representing the number of parallel workers. Acceptable values for this number are:

  • -1, representing synchronous execution (this is the default across InVEST)

  • 0 representing threaded task management

  • Any other positive integer represents the number of processes that will be created to handle tasks. 2*multiprocessing.cpu_count() is usually a good number.

Warning

If you use this feature, you must wrap your script in a if __name__ == '__main__': condition. Failure to do so will result in a fork bomb (https://en.wikipedia.org/wiki/Fork_bomb).

Using the parameter study example, this might look like:

if __name__ == '__main__':
   args['n_workers'] = 4  # Use 4 processes

   #Loops through the values 500, 550, 600, ... 1000
   for threshold_flow_accumulation in range(500, 1001, 50):
       #set the accumulation threshold to the current value in the loop
       args['threshold_flow_accumulation'] = threshold_flow_accumulation
       #set the suffix to be accum### for the current threshold_flow_accumulation
       args['suffix'] = 'accum' + str(threshold_flow_accumulation)
       natcap.invest.ndr.ndr.execute(args)

Internationalization

If you use the InVEST python API to access model names, MODEL_SPEC s, or validation messages, you can translate those strings using gettext:

from natcap.invest import set_locale

# replace with your desired ISO 639-1 language code
# see https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes
set_locale('en')
import natcap.invest.carbon
...

See the GNU gettext manual and the Python gettext documentation for more information.

Summary

The InVEST scripting framework was designed to assist InVEST users in automating batch runs or adding custom functionality to the existing InVEST software suite. Support questions can be directed to the NatCap support forums at http://community.naturalcapitalproject.org.