Model sensitivity analysis using SALib

Clinic at CSDMS Annual Meeting 2019

Description

Interested in which variables influence your model outcome? SALib (Sensitivity Analysis Library) provides commonly used sensitivity analysis methods implemented in a Python programming language package. In this clinic we will use these methods with example models to apportion uncertainty in model output to model variables. We will use models built with the Landlab Earth-surface dynamics framework, but the analyses can be easily adapted for other model software. No prior experience with Landlab or Python is necessary.

Software

Software that we will use includes Anaconda, SALib, Landlab, and Jupyter. The installation instructions below were adapted from www.earthdatascience.org.

Using SALib for sensitivity analyses in general is the focus of this clinic. Additionally, Landlab will be minimally introduced in order to demonstrate how SALib can be used with model results. If you wish to learn more about Landlab in a video before or after the clinic, see this overview: https://csdms.colorado.edu/wiki/Presenters-0407.

Clinic preperation

Please prepare your computer for the clinic by completing the sections, Software installation and Create a conda environment.

Get help

Search the issues or submit a new issue at https://github.com/nathanlyons/sensitivity_analysis_clinic_CSDMS_2019/issues.

Software installation

We will use the Anaconda Python 3 distribution for this clinic. Anaconda is a distribution of Python that comes with many of the scientific computing packages that we will need, including Jupyter.

If you already have Anaconda, verify that you have the 64-bit version of conda installed, and then you can skip to the Create a Conda environment section. You do not need to install Anaconda again if you already have Anaconda for Python 2 setup (although 64-bit Anaconda is required, contact me if you need further explanation). We will be working with Python version 3 in this clinic, but a Python 3 environment can be installed into an Anaconda 2.x distribution. We will create a conda environment with Python 3 later in these instructions.

Windows

IMPORTANT: if you already have a Python installation on your Windows computer, the settings below will replace it with Anaconda as the default Python. Please contact me if you have questions or concerns about this.

Download the Anaconda installer for Windows. Be sure to download the Python 3.7 version!

Run the installer by double-clicking on the downloaded file and follow the steps bellow:

1. Click “Run”.
2. Click on "Next".
3. Click on “I agree”.
4. Leave the selection on “Just me” and click on “Next”.
5. Click on "Next".
6. Select the first option for “Add Anaconda to my PATH environment variable” and also leave the selection on “Register Anaconda as my default Python 3.7”. Click on “Install”.
7. When the install is complete, Click on “Next”.
8. Click on “Skip”.
9. Click on “Finish”.

Test installation

1. Search for and open the Anaconda Prompt program.
2. Next, type conda and hit enter. If you see a list of commands that you can execute, then Anaconda Python has been installed correctly.
3. Close the Terminal by typing exit.

Mac

1. Download the installer: Anaconda installer for Mac. Be sure to download the Python 3.7 version!
2. Install: Anaconda—Double-click the .pkg file.
3. Follow the prompts on the installer screens.
4. If you are unsure about any setting, accept the defaults. You can change them later.
5. To make the changes take effect, close and then re-open your Terminal window.

Test installation

1. Search for and open the Terminal program (found in /Applications/Utilities).
2. Next, type conda and hit enter. If you see a list of commands that you can execute, then Anaconda Python has been installed correctly.
3. Close the Terminal by typing exit.

Linux

1. Download the installer: Anaconda installer for Linux. Be sure to download the Python 3.7 version!
2. In your Terminal window, run: bash Anaconda-latest-Linux-x86_64.sh
3. Follow the prompts on the installer screens.
4. If you are unsure about any setting, accept the defaults. You can change them later.
5. To make the changes take effect, close and then re-open your Terminal window.

Test installation

1. Search for and open the Terminal program.
2. Next, type conda and hit enter. If you see a list of commands that you can execute, then Anaconda Python has been installed correctly.
3. Close the Terminal by typing exit.

Create a Conda environment

First, lets update Conda itself:

conda update conda

And then:

conda update anaconda

To avoid software conflicts, I created an environment specifically for this clinic that contains all of the packages that you will need.

Follow these steps to install the clinic environment:

1. Download the environment file: https://docs.google.com/uc?export=download&id=1_Bqui1JcHPyKs6fUbuKk0jEO4G7XYnsi
2. Open the terminal on your computer (i.e. Anaconda Prompt on Windows or Terminal on a Mac/Linux).
3. In the terminal, navigate to the folder containing enviroment.yml that you downloaded in the previous step (e.g. cd downloads).
4. Then, type in the terminal: conda env create -f environment.yml. This will only work if you run the command in the directory where you placed the environment.yml file.

Note that it takes a bit of time to run this setup, as it needs to download and install each library, and that you need to have internet access for this to run!

You can confirm that the enviroment was created, and that your computer is prepared for the clinic by entering the following command in Terminal (Linux/MacOS) or Anaconda Prompt (Windows):

conda info --env

The environment sa-clinic should be in the list of environments.

Your computer is now ready for the clinic.

About conda environments

Anaconda allows you to have different environments installed on your computer to access different versions of Python and different libraries. Sometimes libraries conflict which causes errors and packages not to work.

For more information about conda environments check out the conda documentation.

Manage Your Conda Environment

Always make sure that the sa-clinic environment is activated before doing work for this clinic.

You can have different Python environments on your computer. Anaconda allows you to easily jump between environments using a set of commands that you run in your terminal.

View a List of All Installed Conda Environments

You can see a list of all installed conda environments by typing:

conda info --env

Activate a Conda Environment

To activate an environment, use the Terminal to navigate to your earth-analytics directory (e.g. cd to the directory). Then, type the following command to activate the environment (e.g. sa-clinic):

conda activate sa-clinic

For older installations of Anaconda (versions prior to 4.6) on Mac, Linux, and Git Bash for Windows, type:

source activate sa-clinic

Windows Users: The lessons on this website assume that Windows users are using Anaconda Prompt as their primary terminal. If you need to activate a conda environment using the Command Prompt, you will need to use the following command: activate sa-clinic

Once the environment is activated, the name of the activated environment will appear in parentheses on the left side of your terminal.

Data Tip: Note that after you restart the Terminal, the sa-clinic environment is no longer active. You will need to activate the sa-clinic environment each time you start the Terminal by running the appropriate command provided above for your operating system.

Deactivate a Conda Environment

You can deactivate a conda environment if needed. Deactivating the environment switches you back to the default environment in the terminal session.

Mac and Linux Instructions:

Source deactivate sa-clinic

Windows Instructions

deactivate sa-clinic

Delete a Conda Environment

If you ever want to delete an envrionment, you must first deactivate that environment and then type:

conda env remove --name myenv

and replace myenv with the name of the environment that you want to remove.

Do not delete your root environment.