Jahnavi Shah

PhD Student, Planetary Science

GIS-based Multi-Criteria Decision Analysis

This blog post is dedicated for the Planetary Science Seminar course I’m taking this semester. I will share a general overview of a tool I recently used for a research project over the summer. The tool/method is called the Multi-Criteria Decision Analysis (MCDA) and its use in relation to GIS.

The GIS-MCDA is an optimization method useful for spatial problems where there are multiple and conflicting interests and stakeholders. This tool is applicable in planetary science mission planning scenarios such as landing site selection. The mission planning process involves operational and scientific criteria, which don’t agree with each other at times. For instance, key areas of scientific interests might be inaccessible by a rover. The GIS-MCDA procedure is summarized in the chart below:

The criteria (includes factors and constraints and are in the form of raster layers in GIS) are assigned weights based on their importance to the objective. There are several methods for determining weight of each factor (but I will not discuss that in this post, perhaps another time). A factor is defined as a criterion that enhances or detracts from the suitability of a specific alternative for the activity under consideration. For example, the distance from rover landing site to a potential sampling target may vary from near (=most suitable) to far (=least suitable). Since factors have different ranges and cannot be directly compared, they are usually standardized before assigning scores/weights. A constraint is a bit more straight-forward and serves to limit the alternatives under consideration. It is an element or feature that representation limitations – an area that is not preferred in any way. For example, a ca rover cannot traverse over a slope greater than 20° and that sets the constraint for the slope criteria. Constraints can be represented fairly easily as a mask with Boolean values (0 or 1). The MCDA combines factors and constraints using the following formula to reach a decision:

ArcGIS tools used for MCDA:

1. Reclassify

The Reclassify tool can be used to create masks by converting layers to Boolean values using a threshold. NoData values can be reclassified to 0 to avoid excluding them when using the raster calculator. This tool can also be used to create criteria ranking.

2. Raster Calculator

The MCDA formula can be implemented directly within a GIS raster calculator:
S = ((F1*0.67) + (F2*0.06) + (F3*0.27)) * cons_boolean
where F1, F2, F3 and cons_boolean are raster layers representing the factors and constraints.
This tool allows pixel-by-pixel operations using one or multiple raster layers. The constraints layers are multiplied with one another, and a weighted sum is used for the criteria. It is possible to use raster calculator with layers of different projections, cell sizes or extents, but it is recommended to use layers with standardized properties.

The last step of the MCDA process is to validate the result. It is important to assess the reliability of the output by:
1. Sensitivity analysis: check how the result is affected by altering the set of criteria and the weights of the factors.
2. Ground truth verification: conduct a field survey to verify sample areas, if possible. This may not be possible in a landing site selection/mission process scenario. In that case, it is useful to create criteria maps and manually analyze the output. It would be useful to perform a local analysis of the area.

The MCDA process can be iterated as criteria evolves during the mission planning process to obtain the results that best reflect the data, operational constraints, and scientific criteria available at the time. This is not an exhaustive list of tools available for GIS-based MCDA, but a simple introduction to a some common functions that I used for my project. The resources listed at the end provide information about more functions and details about the ones presented here.

1. ArcGIS notes by Ryan and Nimick, 2019
2. Lecture notes by Estoque, 2011
3. Eastman et al., 1995

Next Post

Previous Post

Leave a Reply

© 2022 Jahnavi Shah

Theme by Anders Norén