Archival Research

For this project, I collected data from a wide range of sources about the location, number, and types of mounds, as well as the year of first recordation (Supplemental Appendix I). The Iowa Site File paper forms of the Iowa Office of the State Archaeologist (OSA 1961–2003), the OSA digital site forms after 2002, and the archives of significant early Iowa mound researchers were consulted (for the full list of sources, see OSA 2011–2023); a full discussion of historic mound evaluation is included in Whittaker (Reference Whittaker2023). The limitations of collecting and analyzing large volumes of historical mound data were discussed in detail by Arzigian and Stevenson (Reference Arzigian and Stevenson2003:47–56) in their analysis of Minnesota sites. Their observations hold true for the Iowa data: the quality of historical data ranges widely, from detailed survey maps of individual mounds to passing mentions of vaguely recalled possible mound groups made decades after their destruction.

Historical records can provide data both about the shape of mounds expected at a site and their number. For records in which mound shape was not specified, I coded the mound as conical, by far the most common mound shape—with the expectation that if the mounds were unusual they would be noted as such. I tallied sites that are recorded only as “mounds” or “mound group” as two conical mounds, the minimum number of mounds and the most likely mound shape. When a range of mound counts was given, the highest number was counted.

Because of the low numbers of mounds noted between 1840 and 1869, those were combined with the 1870s in my analysis of the decades in which mounds were reported. In the 1840s, 57 mounds were recorded from three sites; in the 1850s, three mounds from two sites; and in the 1860s, 20 mounds from two sites. In contrast, in the 1870s, 355 mounds were recorded from 43 sites.

Mapping and Checking Mound Data

The 2020 statewide lidar survey conformed to then-current US Geological Survey (USGS) 3DEP specifications, with typical vertical accuracy better than 6 cm RMSD_z and typically more than two measurements per m² (US Geological Survey 2019). This level of accuracy and thoroughness exceeds what can be practically done with in-field EDM survey systems (e.g., Whittaker and Tiffany Reference Whittaker and Tiffany2021), and these data far exceeded the quality of the earlier lidar data and even field maps, showing more mounds at far higher resolution and accuracy (Figure 2; Iowa State University Geographic Information Systems Support and Research Facility [ISUGISSRF] 2023; Office of the Chief Information Officer 2023). In addition to hillshade maps, ISUGISSRF (2023) also provides interactive digital elevation models (DEMs), in which any view is shaded by the relative elevation of all 1 m DEM grid points. This means that, as one gets closer to a possible mound, it is possible to see whether the model projects it as a true rise with slopes on all sides. If a rise had discernible slopes on all sides that clearly break with surrounding topography, it was coded as a clear mound. If it was ambiguous, it was coded as a possible mound. Examples of ambiguity include rises with slight slopes in all directions from a center point, which gradually fade into surrounding topography; rises that look badly distorted but occurred where historical records suggested a mound should be; or rises on slopes in which the summit is not clearly discernible from the natural upslope. Clear mounds were also coded by apparent damage: no observed damage, less than half the mound damaged or distorted, or more than half damaged or distorted. Mounds were coded by shape, including conical, linear, effigy, compound, and other. Aboveground earthworks, such as village enclosures, were placed in the “other” category and so were not included in this study. Possible mounds near existing mound site boundaries were considered part of the existing site if they were on the same landform. The results are shown in Supplemental Appendix I.

Figure 2.

Example site 13WB72. Upper: Richard Flanders's 1962 sketch map in the Iowa Site File (OSA 1961–2003); middle: site area in 2008 lidar hillshade; lower: site seen in 2020 lidar hillshade (ISUGISSRF 2023).

Occasionally, lidar imagery revealed more mounds than were originally recorded (Figure 2). Experiments with using raw LAS data to create better DEM maps of sites, following the method discussed by Whittaker (Reference Whittaker2020), proved fruitless, because the quality of USGS 3DEP processing and filtering performed on the 2020 hillshade and DEM defaults was superior to the processing available in common GIS programs like ArcMap and ArcScene.

Mound Distribution

This study identified previously unknown aspects of mound distribution in Iowa (Supplemental Figures 1–5). For example, by mapping all effigy mounds, it detected a geographic break in their forms: all effigy mounds north of Guttenberg were shaped like a bear or raptor, and all those to the south were a mix of other effigy forms (Whittaker and Collins Reference Whittaker and Collins2022).

A second, newly discovered mound distribution pattern emerged in central Iowa. Two clusters of linear mounds were identified: a large cluster of approximately 200 linear mounds along 40 km of the Des Moines River and a smaller cluster of about 65 in seven sites in Hardin County along 11 km of the Iowa River (Whittaker Reference Whittaker2022). Neither of these clusters had been fully recognized in the archaeological literature. Limited archaeological evidence suggests that the Des Moines River linear cluster is associated with the Late Woodland transition to Great Oasis, a regional manifestation of large-scale Plains–Midwest shifts to village life occurring around 1000 BP (Whittaker Reference Whittaker2022).

Calculating Mound Survivorship

Table 1 presents a summary of all the mounds expected based on historical research and how they are seen in lidar. In Iowa, 7,762 mounds were historically recorded, or 0.138 mounds identified per square mile; this figure is nearly identical to the 0.137 mounds identified per square mile in Minnesota, where Arzigian and Stevenson (Reference Arzigian and Stevenson2003:63–64) noted 11,868 mounds. Conical mounds make up the vast majority of mounds in Iowa, both in historically identified (85%) records and as observed in lidar. Although there are significantly fewer linear and compound mounds, they have higher rates of survivorship. Effigy mounds have the lowest rates of survivorship.

Table 1.

Summary of Mounds of Iowa.

I wanted to use mound survivorship to determine the rate at which mounds were lost and then to project this data backward to estimate the number of mounds lost before they had even been noted. Of crucial interest is the rates at which mounds are lost over time. I assumed that survivorship would be exponential, with the rate of loss being rapid during the early decades of non-Native settlement and then decreasing in recent years.

At first, I looked at survivorship by decade (Figure 3; Table 2). Figure 4 presents all possible mounds seen in 2020 lidar by each decade. These data show a trend of increased survivorship over time. However, neither linear (R² = 0.61) nor exponential (R² = 0.64) trendlines are statistically significant. The two trendlines were close to each other, both visually and statistically, suggesting that survivorship is not exponential over time.

Figure 3.

Mounds identified per decade.

Figure 4.

Percent of mounds possibly seen in lidar by decade identified. Linear trendline R² = 0.6102; exponential trendline R² = 0.6359.

Table 2.

Mounds Identified by Decade.

^a Charles Keyes and Ellison Orr did not clearly date their discovery of 36 mound sites, but their work was conducted between 1900 and 1945.

The lack of R² significance is due to the idiosyncrasies of survivorship. Although mound loss is necessarily cumulative over time, within each decade large sites can skew survivorship. For example, the Sny Magill Mound Group (13CT18) was first mapped in 1885 and contains 95 mounds, now preserved as part of Effigy Mounds National Monument. In contrast, the Harpers Ferry Great Mound Group (13AM79) was first noted in 1892 and contained 922 mounds, now largely destroyed. Such cases play havoc with decade-based survivorship trendline significance.

I decided to further combine data to obtain a statistically significant trendline and so determine whether loss was logarithmic. At least three data groupings are needed to determine whether a trendline is an exponential curve or a straight line, and grouping data would presumably even out the decade-to-decade idiosyncrasies. Indeed, there appeared to be natural breaks in the data that allow it to be divided into three roughly even eras (Table 3). Era 1, 1840–1900, is dominated by Lewis's surveys and early mound accounts. Era 2, 1900–1959, was dominated by Keyes and Orr's surveys. Era 3, 1960 to the present, is dominated by modern cultural resource management and lidar investigations. Each group contained roughly equal counts of mounds, and each era spanned about 60 years. This division also allowed the inclusion of 98 undated mounds recorded by Orr and Keyes at uncertain times during Era 2.

Table 3.

Mounds Identified by Era.

^a Includes Keyes-Orr sites from Table 2.

Grouping by era proved to be highly significant, as revealed by both linear (R² = 0.9978) and exponential (R² = 0.9896) trendlines (Figure 5). The trendlines disconfirm my initial assumption regarding survivorship and support the counterintuitive idea that survivorship is linear.

Figure 5.

Percent of mounds possibly seen in lidar by era. Linear trendline R² = 0.9978; exponential trendline R² = 0.9896.