Background & Summary

The Morrow Plots at the University of Illinois Urbana-Champaign holds great historical and scientific significance as the oldest continuously operated agricultural experiment in North America. Initially established in 1876 by professors Manly Miles and George E. Morrow, the Plots investigates the impact of crop rotation on maize (Zea mays L.) yields, the major crop in the eponymous region of the Corn Belt1. In 1904 the Plots was modified to include the first set of fertility treatments, for which there would be four major shifts to reflect several historical changes in nutrient management of United States and global agriculture2. Over time the Morrow Plots has been a site of active agricultural research and serves as a living record of the history of agriculture in Illinois and the wider region. In year 92 of the experiment (1968), the Morrow Plots was deemed a National Historic Landmark3. Although data have widely been reported in many publications over the course of the experiment, a comprehensive digital dataset was first published in 2022 by the Morrow Plots Data Curation Working Group4.

The Morrow Plots experiment was inspired by the Rothamsted plots, which are today the world’s oldest continuous agricultural experiments. Established in Hertfordshire, UK in 1843, the Rothamsted experiments test the impact of soil fertility treatments on a variety of England’s most common grain crops5. In 1875, University of Illinois agricultural chemistry instructor Chas W. Silver visited Rothamsted, which he described as “made with the accuracy of an exact science giving results that will survive the fluctuations of commercial prices, do away with the ignorance of empirical cropping, found an accurate system of utilizing the soil and farm refuse, and be a source of satisfaction and a boon to humanity”6. Upon returning to Illinois, Silver proposed a similar long-term experiment, which responded to urgent local and national needs. In the early years of Illinois agriculture, many farmers believed the prairie soil was so rich that it would support crops indefinitely7,8. However, by 1876 after 40 years of largely unmitigated nutrient mining due to continued cropping without fertilization, initially highly fertile prairie soils were showing signs of exhaustion with maize yields in the state down 50 percent9. At the same time the federal government was well aware of declining fertility and the lack of new lands in the post-bellum US, to the point of sending delegations of researchers to eastern Asia to understand how the same field could be farmed “for forty centuries” without exhausting soil fertility10. The Morrow Plots was designed to address this growing issue through scientific experimentation and was one of the early research efforts that helped establish agriculture as a subject of academic study in the US8,11.

The Morrow Plots was established in 1876 in Urbana, Illinois, well south of the few campus buildings that existed at that time in a field now classified as the Flanagan soil series (fine, smectitic, mesic Aquic Argiudoll)12. Dark colored at the surface depths and somewhat poorly drained at subsurface depths, the soil is representative of the larger US Corn Belt: developed on loess overlying glacial till and under tallgrass prairie. (For further description of the experiment’s soil, climate, and landscape, see the 1997 article from Darmody and Peck13). High organic matter and base saturation in tandem with circumneutral pH and large water holding capacity enables this soil type to support some of the highest crop yields in the world12.

Land use prior to the establishment of the experiment may have impacted initial results of crop rotation on yields. Much of the surrounding area was converted to farmland around the 1840’s, but the conditions of the field at that time are unclear14. The university was founded in 1867, and as it grew around the plots, the name of the experiment evolved. It was first referred to as Experiment 23 on crop rotation and was later known as part of the North Farm, as the Urbana Experiment Field, the Old Experiment Field, the Morrow Plats (“plat” being a variant of “plot”), and finally as the Morrow Plots. They now sit in the center of campus just south of the main quad (40°06′16.4″N 88°13′34.1″W).

Longitudinal agricultural experiments are rare, and enduring ones more so. In 1927, Herbert Mumford, Dean of the College of Agriculture and Director of the Agricultural Experiment Station, wrote that the Morrow Plots experiment is “all the more remarkable because it was done at a time when the Agricultural Experiments Stations of the country were young and the pressure was for investigations that would yield quick returns”15. As that same pressure continues today for different reasons16, the Morrow Plots (1876) remains a rare long-term experiment and is the oldest of three ongoing agricultural experiments begun in the US in the 1800’s, along with the Sanborn Fields in Missouri (1888) and the Magruder Plots in Oklahoma (1892)17. A newer generation of long-term agricultural experiments established in the 20th century is much larger and includes the 18 sites in the US Department of Agriculture’s Long-Term Agroecosystem Research Network18,19.

The data produced by and insights derived from the Morrow Plots over the past nearly 150 years allows researchers to quantify temporal trends in crop yields and soil properties, including but extending beyond fertility. In 1908, after the first 30 years of the experiment, Hopkins, Readhimer, and Eckhardt concluded that applications of manure, limestone and phosphorus (P) could lead to rapid improvements in maize yields, especially in fields with a crop rotation that had not previously included legumes20. This was the basis for Professor Hopkin’s at-the-time revolutionary proposition of a “system of permanent soil fertility”21, which built the foundation for nutrient management in the early 20th century in Illinois and the greater US Corn Belt22. At the 50-year mark in 1927, DeTurk, Bauer, and Smith demonstrated that in order to maintain yields over long periods of time fertilizers need to be used in tandem with crop rotation15. After the first 100 years, the Morrow Plots showed that improved technology such as new maize hybrids and concentrated inputs of nutrients (i.e., synthetic fertilizers) could not only mitigate but even reverse soil nutrient depletion and increase yields markedly23. At the end of the 20th century, in a comprehensive statistical analysis that included historical weather data, Aref and Wander reported maize yields to be significantly correlated not just with weather conditions in the current year but also with total precipitation over the 2 previous years2. More recently, in 2011, Nafziger and Dunker found that losses in soil organic carbon under agriculture (i) begin to stabilize after a century, (ii) are greater with simplified crop rotations and lack of fertilization, and (iii), are influenced by the initial soil organic carbon level24,25.

Currently, the research team managing the experiment is investigating multiple questions that have arisen due to new and enabling technologies and/or emerging questions for the 21st century that the Morrow Plots is uniquely suited to answering. Current and planned research will explore questions about the emergence of antibiotic-resistant microbes in the latter half of the 20th century, the age and source (e.g., original prairie grasses vs maize) of organic matter in the soils, the effects of long-term agriculture on water filtration and storage, and when and how persistent contaminants like microplastics began to appear in soils. Importantly, these ongoing projects are yielding new kinds of data different from the traditional data recorded to date, which have been largely crop yields and agronomic practices. While the Morrow Plots will continue, the University is also building on this legacy with a new experiment, the Alma Mater Plots, a nearly 80-acre research field that will include 64 1.2-acre treatment plots and four replications. Once established, the Alma Mater Plots — conceptually the Morrow Plots V2.0 — will operate for the next 150 years to test the long-term impacts of agriculture practices common today, which will be identified with input from stakeholders throughout the state26.

The Morrow Plots has inherent limitations resulting from small plot size, lack of randomization, and changes to the experiment design over time18. However, these design flaws are trade-offs worth accepting because they allow testing of long-term processes and outcomes simply not accessible to most agricultural experiments. On the other hand, the small size of the Plots, less than two acres total today for all 24 subplots, have made them easier to maintain than some larger experiments that were ended early due to lack of resources such as the University of California Davis Century Experiment27. The Plots has survived long enough to reflect changes in crop breeds, farming practices, technology, and fertility treatments. Although these changes complicate testing specific hypotheses, they capture the impact of major changes in the Illinois farming industry, from its inception through today – a trade-off for any sufficiently long-term agricultural experiment28.

Methods

This section explains how data were acquired and processed and provides experimental detail essential for interpreting the data. First, this section opens with an explanation of how data were acquired from contemporary and historic sources, and the processing required to synthesize these sources into one cohesive dataset. Then, the major phases of the experiment are described, with a focus on the evaluated experimental treatments of crop rotation and soil fertility treatments, and the experimental design such as the replication and arrangement of plots and subplots. Finally, more detail is provided about the complex history of fertility treatments that were updated over time to reflect changing practices in planting, cultivation, and harvesting.

Data sources

The published dataset is the collaborative product of two groups at the University of Illinois Urbana-Champaign: the multi-disciplinary Morrow Plots Data Curation Working Group and the research group of Dr. Andrew Margenot of the Department of Crop Sciences. The Working Group was established in 2018 at the request of the College of Agricultural, Consumer, and Environmental Sciences and includes information professionals from that college as well as the University Library and Archives. The Working Group formed with two goals: (1) locate and preserve historical data records associated with the Plots; and (2) compile, curate, and publish an openly accessible, longitudinal dataset for the entire span of the experiment to date29. The resulting dataset contains newly transcribed data from analog records, digital data that has been shared in previous publications, and digital data from recent years that has not been previously reported. The Working Group collaborated closely with Dr. Andrew Margenot’s research group, which manages all aspects of the experiment and stewards the ongoing digital data record and soil sample archive. Both groups also continuously communicated with University stakeholders, and received approval from University administrators prior to publication18. Going forward, the Working Group will continue to work with Dr. Margenot’s group to update the published dataset as the experiment progresses.

The dataset was compiled from four sources: an archival notebook, a yield table, a tracking spreadsheet, and a soil sample inventory (Table 1). The notebook is preserved in the University Archives and provided early crop varieties and planting dates, which were manually transcribed from the handwritten text30. The yield table was published in a 1984 Agricultural Experiment Station report and contains crop, treatment strategy, and yield data, which were digitized23. Digital surrogates of both source records are openly accessible online. The tracking spreadsheet, which provided much of the detail included in the final dataset, is an internal record maintained by Dr. Margenot’s group. Finally, a soil sample inventory, also from Dr. Margenot’s group, was used to flag years in which soils were sampled from the Plots and added to an archived collection maintained at the University. Each of the four data sources are described below followed by a summary of how these sources were combined and augmented to create the published dataset.

Table 1 Date range for and data acquired from each data source.
Full size table

Archival notebook (1876–1913)

The earliest known surviving data source for the Morrow Plots is a handwritten notebook in the University Archives covering the founding of the experiment in 1876 through 1913 (Fig. 1)30. The first 12 years of the experiment (1876–1887) are summarized with one page of text and a table of crops planted. No explanation is given for the missing data from these years, although other records from this early period share the same gap31,32,33. Later years (1888–1913) include tabular and text data for crops, varieties, planting dates, treatments, and yields. Planting dates and crop varieties were manually transcribed from the notebook’s narrative text into a spreadsheet and imported into RStudio to be combined with data from the other three data sources. Data for crops, treatment strategies, and yields overlap with, and were used to validate data from, the yield table described below.

Fig. 1
figure 1

Archival notebook summary of the early years of the experiment (1876–1888). The annotations in red ink demonstrate past disagreements over the experiment’s start date.

Full size image

Future work could include extracting additional details from the notebook narrative, however, the depth of detail varies from year to year and contains a mix of concrete data and broad generalizations, such as: “The ground was well prepared in the spring and planted to corn in checked rows 3 feet apart each way, 3 kernels to the hill. The corn was later thinned to 2 stalks per hill. The corn had the customary amount and kind of cultivation usually given to corn.”30 At the time such practices were so customary and usual that no further explanation was recorded. The narrative inconsistently reports specifics of fertility treatments such as the amount of input applied. These may be extracted from the narrative accounts in the notebook and added to the dataset in the future but were not a priority for this version.

Yield table (1888–1954)

Many yield tables have been published in print, the most comprehensive of which, in terms of years covered, was included in “The Morrow Plots: A century of learning,” a 1984 report published by the Agricultural Experiment Station23. Table A1 from that report (Fig. 2) includes crops planted, treatment strategies, and yield for the first several decades of the experiment (1888–1954). These data were manually transcribed into a spreadsheet with the same organizational structure as the printed table, checked against the printed table for accuracy, then reorganized into a machine-readable format and checked again before being imported into RStudio and combined with the other data.

Fig. 2
figure 2

An excerpt of the yield table used in the dataset. The full table spans 1888–195423. A digitized version was incorporated into the published dataset.

Full size image

In this table, one annual yield figure was reported per plot until 1904 when each plot was divided into four subplots: two treated subplots (MLrP and MLbP in Fig. 2) and two that received no treatment. From 1904 onward, annual yield for each plot was reported in three figures: one for each of the two treated subplots (MLrP and MLbP), and one average of the two untreated subplots (No treatment). Separate yield rates for the untreated subplots have been reported piecemeal in annual Agricultural Experiment Station bulletins and may be incorporated into a future version of the dataset. Crucially, all yield data are reported at a per acre rate, making it possible to compare yields from entire plots in the early years of the experiment and those from smaller subplots in later years.

Tracking spreadsheet (1955–2021)

The tracking spreadsheet is the primary digital data record for the experiment and is used and stewarded by Dr. Margenot’s research group. This resource includes data that were previously tracked in field notebooks and much of which has been reported in publications, most recently in a 2011 paper that included data through 200924. The tracking spreadsheet is used to record and monitor details on crop planted, crop variety, planting date, planting day, treatment strategy, treatment amounts applied, plant population, yield, the amount of stover removed, and notes about consequential events such as floods and crop damage. Data are recorded as one row per subplot per year and are manually entered by research group members.

The tracking spreadsheet provides more detail than the archival notebook or the yield table but does contain some gaps and inconsistencies. There are two years, 2010 and 2014, for which all data are missing. For the sake of completeness and consistency, rows for these years are included in the published dataset with “NA” values. Other variables contain data for some years and not others. Variety and planting date are missing for most years after 2009. Planting dates are also missing for hay crops in all years, but this is not unexpected since the experiment focuses on maize, and because hay yields are relatively less sensitive to planting date than yields of maize, oats and soybean. Plant population data is available for most maize crops but shows variations in collection methods over time. For the first several decades, population numbers match the planting density rates described in published accounts2,23, but in the last few years more specific population numbers are provided from harvest counts. Moving forward, both planting density rates and harvest counts will be recorded.

On the one hand, population and many other variables that only appear in the tracking spreadsheet provide uneven slivers of data when compared to the whole of the experiment. On the other hand, this is one of the longest running agricultural experiments in the world, and a sliver could cover decades. When reviewing the uneven spreadsheet data, the Data Curation Working Group decided to prioritize detail over consistency. Data are sparse for many of the variables, but even the gaps provide insights into how the experiment changed from one decade to the next. The gaps also represent spaces that are ready to be filled should other records come to light.

Soil sample inventory (1904–2021)

The University maintains an extensive archived collection of soil samples, including samples from the Morrow Plots, dating as far back as 1904. The Morrow Plots samples in the inventory data were cross referenced with the data described above, and used to flag the years for which such samples exist. The inventory dataset is not publicly available at this time but there are plans to publish the developing dataset and ongoing analyses in the future.

Data processing

Each of the three primary data sources (archival notebook, yield table, and tracking spreadsheet) have historical value as records, but their formats significantly limit reuse. With the ultimate goal of publishing a FAIR (Findable, Accessible, Interoperable and Reusable) dataset34, the Data Curation Working Group opted to employ the tidy data format35. The tidy format is interoperable and machine readable with one variable per column, one observation per row, and one value per cell. This format facilitates linking with related datasets, for example, weather conditions or crop price data.

Data were primarily compiled and tidied using R code but required some initial reformatting in Excel prior to import into RStudio (Fig. 3). Each of the three primary data sources presented unique challenges. Although the notebook (1876–1913) has been digitized for access and preservation, none of the text is machine readable. It is entirely handwritten with strikethroughs and annotations from more than one author, and much of the data are embedded in the narrative. Crop, treatment, and yield data from this period are already present in the yield table data (1888–1954) but planting dates and crop varieties had to be manually transcribed from the notebook to a CSV file. The yield table presented a different challenge because the structure of the printed table includes duplicate columns and multiple rows of headers. The data table had to be elongated in Excel before the file could be imported into RStudio. Finally, the tracking spreadsheet (1955–2021), in XLS format, was by far the most complex data file and required the most attention before import. Any Excel formatting not compatible with CSV format, such as color coding and embedded charts, had to be translated or removed. Notes scattered throughout the dataset were condensed into one text column. Formulas were converted to values, column headers were shortened, and column formats were standardized (e.g., dates formatted as dates and text formatted as text). For all data sources, this initial round of cleaning in Excel focused on the basic structure and format of each historical data source and was kept to the minimum amount of processing required before data could be imported into RStudio.

Fig. 3
figure 3

The workflow for cleaning and compiling the dataset included minimal formatting in Excel to prepare data files for import into RStudio where they were cleaned, compiled, tidied, validated, and visualized.

Full size image

Further cleaning in RStudio was necessary to standardize and transform the data before the data sources could be combined. The transcribed notebook data and yield tables were combined using R to create a 1888–1954 dataset. Plant day (of year) was calculated from planting date, when available. The most extensive transformation was in plot divisions. Over time, each plot was divided into eight subplots as new treatment variables were introduced. To allow for longitudinal comparisons, modern plot divisions were imposed backwards in time onto data from all years. For example, in 1888 Plot 3 produced maize at a rate of 54.3 bushels per acre. Although it was undivided at the time, the published dataset includes a separate row for each future subplot (3NA, 3NB, 3NC, etc.), all recorded at 54.3 bushels per acre. These divisions make it possible to make longitudinal comparisons without transforming the data. After subplots were standardized for 1888–1954, this data was then merged in RStudio with the 1955–2021 data from the tracking spreadsheet, which also recorded yield at per acre rates.

Once data for all years was combined in RStudio, several new variables were added and values in existing variables were standardized to facilitate analysis. A “rotation” variable was added to denote whether maize is planted every one, two, or three years. Previously, this either required outside knowledge of the planting schedule or had to be inferred using several years of crop data. Within the “crop” variable, continuous maize was differentiated from maize planted in rotation with other crops. Yield was split into two variables (“yield_bush” and “yield_ton”) to reflect differing units of measurement: tons for hay crops and bushels for all other crops. A “phase” variable was added, and binary variables were included to flag treated plots and years when maize was planted in all plots. The diagram pictured below (Fig. 4) and published in the codebook visualizes the various data sources and maps specific variables from their source to the final compiled dataset.

Fig. 4
figure 4

Morrow Plots data migration map. The center column shows variables in the published dataset with arrows showing connections to historical data sources.

Full size image

The experiment

This section highlights the major shifts that have ensured the experiment remains relevant to regional agriculture over time. These shifts were categorized into five distinct phases by Aref and Wander, and mirror significant changes to the crop species and fertility management practices employed by Illinois farmers (Fig. 5)2. Practices used at the Plots reflect those used on commercial farms, but are usually only introduced long after they have become well established in the region, as is typical of long-term experiments28. For example, by the 1870’s many Illinois farmers had installed tile drainage but it was not introduced to the Plots until the early 1900’s7. Such changes, including currently proposed changes, are carefully planned to protect the Plots from the vagaries of short-term trends.

Fig. 5
figure 5

The 5 phases of the Morrow Plots experiment to date represent shifts in experiment design related to plot divisions, crops, and soil fertility treatments.

Full size image

Phase 1 (1876–1903)

In the first stages of the experiment there were times when early records were forgotten and later resurfaced, which caused confusion about the start date36. In the 1880 Board of Trustees report, Professor George E. Morrow states that the “formal commencement” of the experiment was in 187937. This has often been publicized as the start date but internal records and reports from the late 1800’s show the experiment first began three years earlier in 1876 under Professor Manly Miles30,31,32,33. When Miles left the University later that year, stewardship of the experiment shifted to his replacement, Professor Morrow8.

Some sources indicate that the experiment began with only three plots3,8, but most early records report ten half-acre plots with three crop rotation schedules (Fig. 6). Plots 1–3 were planted in maize every year. Plot 4 alternated between maize and oats (Avena sativa). Plots 5–10 originally followed a 6-year rotation of: “Corn [maize], 2 years; oats, 1 year; meadow, clover, timothy, or both, three years”31. Mammoth and medium red clover (Trifolium pratense) were planted in Phase 1, as well as timothy (Phleum pratense) and an unspecified “grass” for hay production23,30. Records of yield data extend back to 1888 when the Agricultural Experiment Station was established on campus30. In an 1885 interview with The Prairie Farmer, a weekly agricultural publication, Professor Morrow claimed early yields were tracked38. If such records were kept, they have since been lost. In both published and internal records, the earliest recorded yield data are from 1888.

Fig. 6
figure 6

Divisions, crops, and treatments over time for the three enduring crop rotations of the Morrow Plots. Crops grown and treatments applied changed at key phases of the experiment to reflect evolving farming practices in Illinois. Plots 3, 4, and 5 follow a one-, two-, and three-year crop rotation, respectively. Hay crops have varied but were primarily clover until 1952 and alfalfa thereafter. The plots were subdivided in Phases 2 and 3 as new treatment variables were introduced. Subplots were initially designated by direction (Northwest, Northeast, etc.). Since Phase 3, each plot has been divided into four Northern subplots (NA, NB, NC, and ND) and four Southern subplots (SA, SB, SC, and SD).

Full size image

During Phase 1, Plots 1 and 2 were the only two plots that received any soil fertility treatments. Plot 1 received stable manure annually at a rate of 20–28 tons32 or “24 two-horse wagon loads” per acre33. Plot 2 received a variety of commercial fertilizers according to no recorded pattern. These fertilizers included dissolved bone-black, muriate of potash, sulfate of potash, sulfate of ammonia, superphosphate, ground bone, and cattle and hog waste tankage30,31,32,33. In 1895 Plots 1 and 2 were discontinued and the land used as a site for a new observatory, which still stands today.

At the turn of the century the 6-year crop rotation schedule in Plots 5–10 began to break down. In 1900 Plots 5–10 were used for oat variety trials30. In 1901 oat trials continued in Plots 6–10, but Plot 5 shifted to a 3-year rotation of maize, oats, and hay30. Since then, every 6 years maize synchronizes across Plots 3–5, allowing for a direct comparison across plots grown in the same season. In 1902 Plots 6–10 were phased out of the experiment to make space for future campus construction projects, leaving Plots 3, 4, and 5, which still endure today30.

Phase 2 (1904–1954)

Phase 2 brought significant changes to the structure of the experiment, and the first use of fertility treatments on enduring Plots 3, 4, and 5. In 1904 each of these three plots was divided into four subplots to accommodate the introduction of two new treatments that followed the then new Illinois system of permanent soil fertility designed to manage acidity with limestone and replenish nutrients extracted from the soil by crops with a mixture of manure and phosphate inputs21. The southwestern subplots received a combination of manure, limestone, and rock phosphate (MLrP), while southeastern subplots received manure, limestone, and bonemeal as a source of phosphate (MLbP).

Stable or dairy manure was applied to all southern subplots at a rate of 2 tons per acre per year for the first 5 years, and in 1909 the strategy was shifted so the amount of manure applied equaled the mass of crop biomass removed, including grain and residues15. While today the amount of manure or fertilizer added is based on the amount of nutrients, namely P and potassium (K) and sometimes nitrogen (N), removed with crop biomass removal, at the time methods for quantifying nutrient export with harvest and mono-nutrient input sources were either not available or too costly. All southern subplots received four applications of limestone: 0.85 tons per acre in 1904, 5 tons per acre in 1919, 3 tons per acre in 1943, and 2 tons per acre in 194915,23. The phosphate application schedule varied but by 1925 when phosphate applications were paused, the southwestern subplots had received 13,200 lbs of rock phosphate per acre, and the southeastern subplots had received 3,300 lbs of bonemeal per acre15,39. Although the two sources of phosphate were applied in a ratio of 4:1, the per acre cost would have been about the same, providing a useful cost comparison for farmers15. The P content of rock phosphate and bonemeal are unknown, requiring estimates to enable calculation of application rates (i.e., kg P ha−1) and introducing uncertainty into mass balances40.

To help isolate these new treatment variables, at the beginning of Phase 2, walkways 2.5 meters wide were installed between subplots to create buffers between treatments. At the same time, the outer edge of the field was converted to a sod grass border. These changes reduced the total cultivated area of the three enduring plots from 1.5 to 0.75 acres. Further changes were made underground. In the spring of 1904, to improve drainage, agriculture students helped install 4-inch diameter clay tiles under the center walkways crosscutting each plot from west to east20,41. Starting in the late 1920’s, the tile lines appear in plot maps as dashed lines between northern and southern subplots (Fig. 7)15,42.

Fig. 7
figure 7

Phase 2 Map of the Morrow Plots with Tile Lines. The tile lines installed in 1904 are represented by dashed lines between the subplots (15).

Full size image

Phase 2 crop rotations remained consistent with the one, two, and three-year schedules in place at the end of Phase 1, but there were some shifts in the maize varieties and hay crops used. In 1937, maize varieties were changed from open-pollinated forms to newly developed hybrids to reflect changes to Illinois farming. For most of Phase 2 hay included a variety of legumes, primarily red clover (Trifolium pratense), although occasionally cowpeas (Vigna unguiculata) and soybeans (Glycine max) were also used as hay23. In 1953 alfalfa (Medicago sativa) was introduced as an alternative to clover23. In the data, the specific hay crop is sometimes provided, but most years were recorded simply as hay, which complicates interpretation of yields since hay yields depend on the species.

Phase 3 (1955–1966)

Phase 3 marked additional changes in plot divisions and the introduction of then-newly emerged synthetic fertilizers to previously treated and untreated subplots. In 1955, an all-maize year, each plot was further divided into eight subplots – four northern subplots (NA, NB, NC, ND) and four southern subplots (SA, SB, SC, SD) – to accommodate additional treatment variables. The B subplots received a new treatment consisting of highly concentrated and highly water-soluble nutrients. N was applied as urea, P as triple superphosphate, and K as muriate of potash (NsPK)2. The previously untreated NB subplots received the new NsPK treatment plus two applications of limestone in 1955 and 1963 (0LNsPK)23. SB subplots, which had previously been in MLrP, received NsPK and manure, plus two applications of limestone in 1955 and 1963 (MLrPNsPK)23. The other southern subplots (SA, SC, and SD) continued with their MLrP and MLbP treatment strategies: manure applications continued at a rate equal to the dry mass of crops removed, and limestone was applied once in 1955 at a rate of 2 tons per acre23.

Phase 4 (1968–1997)

In 1968 the crop rotation in Plot 4 was changed from maize-oats to maize-soybeans. Starting in the early 1950’s, soybeans surpassed oats as Illinois’ second most important crop after maize in terms of acreage and revenue43. In terms of treatments, Phase 4 entailed some simplifications in tandem with a new high-intensity treatment, and the use of soil tests, which had been recently developed – for the first time in history – for P at University of Illinois Urbana-Champaign as the Bray P1 test44. SC and SD subplots continued their MLbP treatment, but the strategy name was simplified to MLP. They continued to receive manure applications but received no new applications of limestone or phosphate23. Treatment strategies were also simplified for NB and SB, the two subplots where commercial fertilizers were first introduced in Phase 3. These previously had slightly different strategies (0LNsPK and MLrPNsPK), but were both streamlined to limestone, N as urea, P as triple superphosphate, and K as muriate of potash (LNPK)2. SA subplots received a similar treatment but with higher rates of N (LHNPK). Applications rates and schedules for LNPK and LHNPK were determined by soil tests of Bray P1 for P and ammonium acetate for K2.

Phase 5 (1998-present)

During the fifth and most recent phase of the experiment, the crop rotations remained consistent. The only major change to treatments was that the high-intensity LHNPK treatment in SA subplots was discontinued and replaced with LNPK. The three LNPK subplots (NB, SA, and SB) receive N as urea, P as diammonium phosphate (though it contains N, it is the most common P fertilizer still used today in Illinois), and potassium chloride. Amounts of P and K are determined by soil tests. MLP subplots (SC and SD) continue to receive manure equal to the dry weight of crops removed.

Currently, the MLP subplots (SC and SD) are fertilized with solid dairy manure applied in the fall preceding the maize crop. Manure is typically broadcast by hand and incorporated by discing. However, in 2011, 2013, and 2015 liquid manure, including swine manure, was injected as a slurry and measured in gallons. In the spring, nutrient rates are calculated and pre-measured for the LNPK subplots (NB, SA, and SB). Fertilizers are broadcast using a push spreader. Plot boundaries were confirmed as recently as 2020 by using a metal detector to locate the metal marking pins embedded around the perimeter of the plots at the ends of each walkway. The per-acre rate of each nutrient for synthetic fertilizers actually applied is recorded in a spreadsheet. The weight of manure is also recorded, although the exact nutrient content, including accounting for residual moisture and thus dry manure mass and nutrient rates (kg ha−1) applied, was not measured until 2021. Beginning in 2021, grain and hay samples were also analyzed for nutrient contents so that nutrient export rates with harvest could be recorded while also providing information on crop nutrient density.

Fertility treatments

Given the purpose of the experiment and its long history, the treatment data is fairly complex. Each phase tests multiple treatment strategies, and each strategy employs multiple treatments (Fig. 6), which are applied in varying amounts and according to different schedules. To handle this complexity in the dataset, treatments are represented across multiple variables. At the most basic level, a True/False “treated” variable flags a subplot as treated or untreated. A “treatment” variable denotes the overall treatment strategy employed for a particular subplot. Over the course of the experiment eight different treatment strategies have been employed (untreated, MLrP, MLbP, MLP, 0LNsPK, MLrPNsPK, LNPK, and LHNPK), and the number and type of strategies used has changed with each phase (Fig. 8).

Fig. 8
figure 8

Treatment strategy history. The number and type of treatment strategies changed with each phase. Treatments have been introduced to previously untreated subplots twice: in Phase 2 (1904) and in Phase 3 (1955).

Full size image

Where possible, the dataset also includes specific treatment amounts applied each year, but these data are sparse for three reasons: (1) some treatments are not intended to be applied every year or are applied only on an as needed basis; (2) some treatment amount data points are missing; and (3) some data were recorded by weight and others by volume. For all these reasons, the “treatment” variable noting the strategy (e.g., MLrP, LNPK) is a much better indicator of treatment than the amount applied in a particular year. For the years before 1955, treatment amount data are completely missing from the dataset. Wherever possible, amounts for these years have been mined from publications and provided in the phase descriptions above. After 1955, the specific amounts applied are recorded in the dataset as tons per acre for limestone, as lbs. per plot for manure, and as lbs. per acre for synthetic fertilizers. Three times in Phase 5 manure was injected as a slurry, necessitating its measurement in gallons and requires assuming a certain density of manure solids to convert to solid mass per acre. Details of the slurry applications are retained in the dataset as notes. Future work could involve converting the slurry data to solid mass per acre measures, as well as mining pre-1955 treatment amounts from published records and incorporating them into a new version of the dataset.

In addition to the prescribed treatment strategies, the basis for crop residues returned to the soil has also varied. In Phases 1 and 2 (1876–1954) all crop residues were removed and only stubble and root residues remained23,30. In Phase 3 (1955–1967), crop residues were returned to the NsPK subplots (NB and SB) except for subplot 5SB23. It is unclear why 5SB was treated differently. Since 1967 all residues have been returned to all subplots2. Green manures were also sometimes planted before maize and plowed under as a fertility treatment. From 1904 to 1966 green manures of alfalfa or white sweet clover (Melilotus albus) were planted in the southern subplots of Plot 4 between crops of oats and maize23. Catch crops are not recorded in the dataset.

Other farming practices

Although plowing and planting methods changed little over the first four phases, plant populations have been adjusted multiple times. Until 1996 most planting was done by hand with maize seeded in hills in rows 40 inches apart2. Since then, most plants are seeded in rows at higher densities. Currently maize and soybeans are planted with 30 inch spacing, oats are planted 15 inches apart, and alfalfa is broadcast. Plowing was also consistent in Phases 1 through 4. Moldboard plows were used for most of the experiment but since 1998 a twisted-shank chisel plow has been used instead in keeping with common tilling practices24. Populations were adjusted earlier and more often, starting in Phase 3. In 1957 maize planting populations were increased, with untreated plots seeded at 8,000 plants per acre, and treated subplots seeded at 12,000–16,000 plants per acre23. In Phase 4 (1968), population rates in some treated subplots were increased to 24,000. In Phase 5 (1998), these were increased to 28,000. In 2013, all maize population rates were increased to 32,000, and some surpassed that in 2021 and 2022 to reflect the Illinois and US-wide trend of increasing density of modern maize hybrids.

Pest mitigation practices have varied over time. Weeds would have been controlled through cultivation at least until the 1930’s when weed science emerged as a topic of study at the University45. Later herbicides were used2. The local fauna has also posed a consistent threat. In 1929 when a new city ordinance prohibited shooting, an employee was hired to drive hungry birds away from the ripening oats while traps were also used to catch up to 40 birds per day46. At the same time, paper sacks were used to protect ears of maize before harvesting47. This practice lasted at least through 1951 and was intended to protect the crops from birds as well as squirrels48. Damage from squirrels is an ongoing issue, perhaps exacerbated by the evergreen hedge currently surrounding the plots. In 2019 and 2021 animals destroyed the maize in multiple subplots resulting in zero-yield harvests. The “damage” variable in the dataset notes the cause of significant damage when known, but this record is undoubtedly incomplete.

Yield is calculated on a per-acre basis using harvest counts, harvested row length and row spacing. Until 2019 maize was harvested and counted by hand. The middle two rows of each subplot were hand-harvested, starting six inches from each end to eliminate potential edge-of-field effects. A combine was then used to harvest the remaining grain. Since 2019 crops are harvested using a 2-row “small plot” combine. An approximately 2.5-meter swath is harvested around the perimeter of each plot and discarded to eliminate potential edge-of-field effects. The remaining interior plot area is harvested to remove grain or biomass, and the yield tallied automatically for maize, soybeans, and oats. Alfalfa is bale harvested using a two-wheel tractor with a sickle cutter attachment. Cut alfalfa is bagged and weighed by hand, as baling is not feasible at the scale of the subplots. From one harvest to the next, the history of these practices provides essential context for interpreting the dataset.

Data Records

The dataset (1888–2021) is available at the Illinois Data Bank, the institutional data repository for the University of Illinois Urbana-Champaign49. Data for available years is stored in a single table where each year is represented by 24 rows of data, one row for each of the eight subplots in three plots, over 134 years for a total of 3,216 rows. Each row records data across 26 variables describing aspects of time, location, crop, yield, and treatment. Variables for each category are listed below with brief explanations for any that are not implicit.

  • Time is represented by five variables: phase, year, all_corn, plant_date, and plant_day. The True/False “all_corn” variable flags years when maize was grown in all three plots, which occurs every six years. The “plant_day” value is the nth day of the year crops were planted calculated from “plant_date” data.

  • Location is represented by three variables: plot, plot_num, and plot_dir. The “plot” variable is the primary identifier and specifies the plot number and sublot directional quadrant. Plot number (“plot_num”) and direction (“plot_dir”) are also included as separate variables to facilitate groupings common in analysis.

  • Crop and yield are represented by seven variables: crop, variety, corn, rotation, yield_bush, yield_ton, and damage. “Crop” is coded as A = alfalfa, C = maize in rotation, CC = continuous maize, H = hay, O = oats, and S = Soybeans. The True/False “corn” variable allows for easy grouping of maize in rotation and continuous maize. Yield is divided into two variables according to unit of measurement. Alfalfa and other hay crops are measured in tons and all other crops are measured in bushels.

  • Treatment is represented by seven variables: treated, treatment, manure, lime, nit, p205, and k20. The True/False “treated” variable separates treated and untreated plots while “treatment” names the treatment strategy employed. All other treatment variables record specific treatment amounts applied when known.

  • Other variables include: notes, stover, population, and soil_sample. The True/False “soil_sample” variable flags those years and plots for which there is a corresponding soil sample in the University’s collection.

Documentation includes a ReadMe file, codebook, and the repository record itself, which contains links to several related resources such as the digitized archival notebook, library resources about the plots, an infographic visualizing the history of the experiment, and the GitHub code repository discussed below. Going forward the Illinois Data Bank will track publications citing the dataset as they appear and link to them in the data publication record. As the experiment is still ongoing, the Data Curation Working Group will periodically add updates to the dataset using the Data Bank’s versioning functionality. The Data Bank includes a version number in each DOI (e.g., 10.13012/B2IDB-7865141_V2) and links all versions together. Older versions will remain available in the Data Bank with a banner directing viewers to the latest version.

Technical Validation

This project presented challenges typically found in historical data sources50,51. Management of the Plots has changed hands many times over the years, making it difficult to track the provenance of the various data sources that make up the published dataset. Furthermore, most of the available data sources contain secondary data. The archival notebook appears to have been recopied from previous records, and also contains strikethroughs and edits in another hand. The yield table was transcribed from print reports, and many of the years recorded in the tracking spreadsheet predate Excel and must have been copied from elsewhere.

Several overlapping strategies were employed to overcome these challenges and ensure the quality and integrity of the published dataset. Where data sources overlapped, values were cross-checked against each other to ensure consistency. Likewise, yield values in the data were validated against values published in Agricultural Experiment Station bulletins. A series of exploratory data visualizations (available in GitHub) were also used throughout the process to evaluate the consistency and completeness of the data and check for missing, duplicated, or contradictory values. For example, in the early stages bar charts checked for completeness across plots and years, and in later stages treatment timelines and yield charts confirmed that the data matched published results. The planting and treatment data align with known crop rotation and treatment schedules with very few exceptions that could be accurate recordings of errors. The validation process revealed issues with 0.2 percent of rows in the dataset. In six instances treatment values were present for untreated subplots. Since these were present in the original tracking spreadsheet, they were retained in the published dataset and flagged in the notes field. Additionally, prior to publication in the Illinois Data Bank, all files were independently reviewed by a professional curator who assessed the deposit and cross-checked the data file with the documentation to ensure any inconsistencies were resolved before release.

Usage Notes

The tidy format and CSV file type make this dataset highly versatile and amenable to many tabular data analysis tools. The Working Group has also published an R data package titled “morrowplots” that allows RStudio users to easily import the data. The R package’s help documentation includes two “vignettes” that provide documented code for: (1) analysing maize yield trends over time, and (2) comparing the data to national yield data from the National Agricultural Statistics Service52. The package and vignettes are also available via GitHub53. As a historical dataset, there are many potential uses for both research and educational purposes. Those interested could explore, for example, integration of historical market prices also available from the National Agricultural Statistics Service or historical weather data available from the Illinois State Water Survey. One note about terminology: while this article refers to the primary crop as “maize,” the published dataset uses “corn” since that is how the data were originally recorded.