Are Your Maps Lying to You? Let’s Talk About Subjectivity in Maps and Geospatial Information
Written by Madeline Lisaius · Follow ·
Imagine you’ve started a new job at an organization that wants to track and manage the spread of a new variant of wheat rust — a crop disease with the potential to destroy an entire harvest. Your role is to apply the organization’s satellite imagery-based algorithm to a new region of small scale and subsistence farms to create a useful map to guide management. To do this, you must first identify where exactly wheat is growing and test to see if this wheat is infected. Additionally, you need to collect some additional useful information to add to the final map and then visualize this all in a clear way. Easy, right?
There are three main methods you personally know to gather knowledge about the region: Through lived experience, a bird’s eye view, and satellite imagery.
Source of Geospatial Knowledge
The on-the-ground, lived perspective is the most intuitive of geospatial information collection. We could walk around fields and document where wheat is and which fields are infected. We might talk with farmers who know why neighboring empty fields were infected in the previous season and more on this year’s high risk factors. We might be shown a community center where farmers regularly gather to discuss problems and learn management approaches. We might ask a group of farmers to draw their own maps of the region, and from this, show networks of knowledge, and seed and machinery sharing. This method of information collection is perfect for finding details, but takes a very very long time.
Description: A farmer describing their crop in the field.
The second method to create geospatial knowledge is through bird’s eye -type photos taken from vantage points like towers, hills, drones, and airplanes. With these types of images, we can see details like patterns of crop cultivation, irrigation, and infrastructure layout. With a trained eye, we might be able to pick out the wheat fields and building types as well — but with limited accuracy. Using solely this information, we can map out a considerable portion of the landscape and its patterns relatively quickly, but with reduced accuracy and detail relative to the on-the-ground method.
Description: A farmer describes the area from a hilltop perspective.
The third and common method of knowledge generation is using satellite-based imagery, which is collected at hundreds of miles from the Earth’s surface in both visible and non-visible spectra. Some publicly available satellites exist which can capture a resolution of about one square foot (30 cm) on earth per pixel, but the most commonly used (and free!) capture about 30 square feet (10 m2) per pixel. This type of data is great for seeing and contextualizing landscape level information. For our work, we could use the multispectral information to quickly identify wheat fields and to a limited extent, unhealthy fields, clusters of buildings, and infrastructure. Similar to the aerial imagery, we might struggle to contextualize some of the information we may be finding.
Description: Farmland data captured by satellite.
The method we choose to use depends on a handful of factors: our time and money, the people we choose to talk to and ask advice from, and even the time of year. Likely, we will have to use some of all three.
Interpretation of Geospatial Information
Now that you have information about the region, you must interpret it. There is surely too much to put all on one map (it would be a mess of colors and lines!) so you have to make decisions about what matters and what is less important.
For the step of creating the base map of wheat and diseased fields, it would be ideal to use the on the ground method to get the most detail, but the resources needed are prohibitive. Our best bet is to use some combination of approaches.
Most approaches use on the ground knowledge and satellite imagery combined. The satellite imagery can give us a general map of wheat and areas of potentially diseased crops, and on the ground information helps us contextualize, identify outliers, and gather invisible spatial information like family ties and centers of community. Even with the algorithms to classify areas of potential wheat disease, we can tune them to either classify densely unhealthy areas with high confidence or to identify outlier “hot spot” areas with high confidence — but not both. We must be careful on what we prioritize as we choose who to talk to (and who not to) as well as the way we prepare our algorithms.
Contextualizing our Map
Now that we have a map of diseased and healthy crops, what additional information should you consider overlaying? And what is the best way to translate this information to map form? Here are some helpful questions to ask yourself:
- Should you prioritize communities that are most marginalized, even if in the minority?
- Should you prioritize areas of most disease, even if preventative management in lower disease presence regions might be a more cost-effective approach?
- Would it be more prudent to prioritize densely populated regions to reduce localized food insecurity, at the risk of overlooking remote and vulnerable families?
- Did you consider cultural demarcations or histories of tension in the region? If so, how do you weigh information about different groups?
- In how you visualize the areas of disease, do we quantify uncertainty in any way? How?
There are no easy or “right” answers to any of these questions. From the data collection to the visualization, this “job” has many subjective decisions. The resulting map — and any map — is from one singular perspective, which is situated in a particular way of viewing the world and understanding a space.
This is what scholar Donna Haraway describes as “partial and situated knowledge” and is central to the creation of maps. Haraway describes knowledge as always coming from a specific point of view (and therefore, is situated in that perspective), and is incomplete (partial), which impacts the production, interpretation, and visualization of geospatial knowledge. Each time we read a map, we are experiencing dozens and dozens of layers of situated decisions that each prioritize specific views.
Stated more concisely, objectivity in mapping is an illusion.
Partial and Situated Knowledge in Mapmaking
The consequences of this partiality are multiple. The first is that perspectives and real lived experiences are both highlighted and erased, whether unintentionally or maliciously.
Second, maps rarely quantify errors or specify their positioning, and culturally we are not accustomed to understanding multidimensional error. We are used to seeing maps as carrying ultimate truth, which isn’t the case — we are almost certainly bound to mis-read the perspective of the map if it’s not contextualized for us, further complicating how we understand the space and time described.
On the flip side, we can re-understand maps as offering a window through which to see the world through a new point of view.
Imagine we see the map (or maps!) that we create as tools for a specific purpose — in this case, for disease management, with the assumptions and decisions made clearly listed. Imagine that there are many different maps of the same region from many different perspectives. The options, when seeing maps as partial and situated, are endless.
Image Description: When you first saw these maps, what did you imagine they described? In this case, these two simple maps show the same wheat growing area which is at risk of infection by wheat rust, but which prioritize different types of information. On the left, the map describes the perspective that prioritizes women farmers — roads that are well-lit at night and the community center that women tend to gather. On the right side, the perspective prioritizes the highest yielding fields and the transport routes used for harvest machinery. Both are valid perspectives, but could lead the decision makers to different conclusions about possible interventions in the region. Combined, they tell a more complete picture of the complex social and economic dynamics in the region.
So the next time you read a map, consider: What intention or lack of intention went into the creation of this map? What perspectives are prioritized or erased in how this information is visualized? How might this map be lying to you? Or alternatively, what unique point of view can we understand with this map?
Blog Contributors
The Rockefeller Foundation