Look to the Land: Visualizing Change in Agriculture
Information is history. Organizations such as Ancestry.com, The Internet Archive, and Google apparently have an insatiable appetite for our historical books, maps, and government documents. Soon the data within these documents will be treated as objects, split into packets, and linked by metadata including author, publication date, keywords, and citations. And just as Ancestry.com can trace the biographies of a staggering number of individuals using linked demographic data, we may soon be able to assemble the biographies of places using geolocated information in agricultural censuses, photographs, mailing lists, and aerial remote sensing. Get ready to open platforms like Google Earth and use time-enabled search tools to access everything from historical street views to agricultural data and literary references. In fact, it’s already in place for certain collections of historical maps, satellite imagery, and aerial photographs. The data floodgates are just beginning to open.
This is worth discussing because although the digital humanities have begun building the tools to help us analyze textual records, spatial data require different approaches. From conservation to agriculture, environmental historians and historical geographers use a range of spatial sources to “look to the land,” a phrase inspired by Wisconsin’s own student of sustainable agriculture, Franklin Hiram King. We can learn from other digital humanists as they design new ways of reading historical landscapes. “Distant reading” is not just a way to generalize textual sources; environmental historians have been doing some variation of this with spatial data for decades. Most scholars who have used Ancestry.com or other powerful “big data” resources, such as a digitized census database or literary corpus, know that there is still no way to automate research and writing. These tools mobilize a panoply of new sources, and they offer the potential for “distant reading,” but the careful interpretation of archival sources is now more important than ever.
I love data visualization. Virtually all of my projects use some form of it, particularly digitally enhanced techniques such as historical Geographic Information Systems (GIS). However, data visualization is rapidly changing with the rest of the digital landscape, and some definitions and disclaimers are in order.
Visualizing data graphically has two main purposes: the first is to convey information—often quantitative, and often in the form of a chart or map. The purpose of these visualizations is to discover patterns and communicate analytical results. They usually require either some knowledge of data architecture or, at the very least, a thoughtful consideration of the method and meaning behind your data (see Minard’s 1869 diagram of Napoleon’s invasion of Russia, below). The second purpose is to present visualizations as an alternative to other images typically produced by graphic designers. Skills in the latter are clearly important in many communications careers, and the measure of a good visualization is often whether it contains innovative graphic design and provides effective click-bait. These are often the very evocative images we see featured in “best data visualization” compilations. A third and relatively new purpose includes data visualization as reconnaissance and primary source discovery. This is one of the design goals of the Trading Consequences visualization tool, for example.
Data entry, processing, and visualization is not for everyone. I used to imagine that visualizations would always yield overviews and conclusions, but I have since learned not to enter data for data’s sake. Too often I found myself trying to extracts stories from data like a dynamite fisher emptying a river, when instead they were best explored by casting a fly rod, discriminately, repetitively, until the river’s choicest narrative appeared. Statistical visualizations can easily become a large investment with strikingly diminishing returns. A judicious historian will know which sources address specific research questions and how to interpret the results.
We assume both that our data are good, and that they tell us something meaningful. But our data are always constructed, partial, and incomplete. This may seem self-evident, but it is critical to realize this before examining a dataset. In fact historical datasets are imperfect compilations of incomplete summaries based on aggregations of very limited questions posed to skeptical and illiterate operators. But often it’s all we’ve got! For example, in the 19th century, the census was the only record most people left. The census tells the story of people, but—critically for environmental historians—it also tells the story of the food, fiber, fuel, and other goods produced in these places. It tells us what agro-ecosystems were capable of under various management practices and energy regimes.
Agriculture will certainly change in the 21st century and forecasting the food system is no easy task thanks to climate change, extended droughts, and unforeseen energy costs. Recent data visualizations help communicate the complexity of the current agricultural outlook. This map shows that almost half of the calories produced by world agriculture is not consumed directly by people.
Another chart, following the Minard design (also called Sankey charts), suggests that these indirect commodities—particularly dairy and meat production—shape most of the ecological footprint of food consumption. Environmental historians must continue to ask: how did consumption choices influence land use in the past, and how did agriculture respond to new environmental and economic pressures?
Clearly the answers are complex, but much of the history of modern agriculture is about the transition to protein-rich diets and energy-intensive meat and dairy production on farms. One region that helps illustrate this is in Prince Edward Island (PEI), an eastern Canadian province with an economy based primarily on agriculture and forestry.
PEI’s land use history shows us how agriculture evolved, adapted, and adjusted, before transforming rapidly in the late 20th century. This island province is best known for its potatoes, and the crop has caused significant environmental degradation. However, this phenomenon, and the modernization it required, was very recent compared with other agricultural regions in North America and Europe. The long-run census data show that PEI’s ruminant system of mixed husbandry was the dominant determinant of land cover, and the potato did not have a significant effect on land use before 1981.
Each of the “bubbles” in this motion chart displays land use and livestock change over time in PEI. As the bubbles move to the right they represent an intensification of land use in the early 20th century, followed by a period of outmigration and farm abandonment. Then, in the Post-WWII period, we see certain areas expanding both agricultural land use and livestock production, particularly in East Prince (green) and Queens Counties (yellow). Perhaps surprisingly, the farms in these areas remain relatively small, as represented by the size of the circles. The most aggressive process of farm consolidation and the abandonment of marginal land occurred in Kings county (blue), not the potato belt of East Prince.
As the chart demonstrates, the largest displacement of mixed animal husbandry on PEI agriculture occurred after 1961, and very unevenly across the province’s 67 townships. This is because the island followed a trend similar to other parts of eastern Canada, where modernization meant a sharp decline in the number of farms and consolidation among a number of larger landowners, and abandonment in many of the outlying regions.
The idea for these charts came from Hans Rosling’s GapMinder project. Rosling uses motion bubble charts to great effect in his presentations. The charts are useful for demonstrating trends in data that contain 3 or 4 variables and change over time. But what good is it? A well organized table will convey similar information. But the motion chart gives a sense of, well, motion. It also allows us to scale-up the analysis for much larger areas; we are currently using the technique to explore Great Plains datasets at the University of Saskatchewan. The chart format allows us to see groupings that represent regional variation within the province. It is a new way to “look to the land,” and it is only one example of new ways to communicate environmental change through large amounts of data.
Data visualizations and long-run temporal analyses have great promise for environmental historians—particularly as the amount of spatial data proliferates around us. But we must also be cautious. Historians have been blinded by data before. When all one has is the census one begins to ask a certain set of questions. As we are seeing with the problems of climate change, we must use a range of sources and allow complex and dynamic analysis. What are we missing that future generations will need to know about our food production systems?
Featured Image: An enumerator visits a farmer for the 1940 US Census. One of the fifty questions Americans were asked in 1940 was, “Does the person’s household live on a farm?” Credit: Library of Congress, LC-USZ62-91199.
Josh MacFadyen is a Postdoctoral Fellow at the University of Saskatchewan, where he examines energy and sustainability in Canadian agriculture. His research uses and develops digital tools such as Geographic Information Systems (GIS) which help historians examine census, remote sensing, and other historical datasets.He works with groups like the Sustainable Farm Systems project and the Historical GIS lab in Saskatoon where he received the data entry assistance of Laura Larsen and Chris Marsh. Josh is also an editor with the Network in Canadian History & Environment (NiCHE). Website. Contact.