1 Introduction
Climate-oriented archaeology has recently used concepts like resilience to understand why and how people’s ability to cope with stress changed over time, which has highlighted the value of interrogating middle-range connections between climate stress and social change at multiple scales (Burke et al. 2021; Degroot et al. 2021; Hussain & Riede 2020; Jackson, Dugmore & Riede 2017; Jones & Britton 2019; Nelson et al. 2016; Scheffer et al. 2021; Spielmann et al. 2016). Regional relationships between people and climate are important but experiences of stress and resilience play out locally. Therefore, downscaling paleoclimate data to spatial and temporal resolutions where climate conditions directly intersect with human decision-making is critical (Contreras et al. 2018).
The need to link human-climate dynamics at human scales is apparent in many archaeological contexts (e.g., Contreras et al. 2018; d’Alpoim Guedes et al. 2016; Degroot et al. 2022; Erickson 2000; Haywood et al. 2019; Hudson et al. 2012; Kohler & Varien 2012; Mamani Pati, Clay & Smeltekop 2011). For example, demographic transitions in the Norse occupation of Greenland and in the Lowland Maya region were once thought to have been deterministically linked to the subsistence and economic consequences of climate change (Degroot et al. 2021; Hussain & Riede 2020). However, in both contexts, higher resolution settlement histories and paleoclimate models have shown that responses to climate were diverse and nonlinear, causing researchers to look for new, multi-causal explanations (Degroot et al. 2021; Dugmore, Keller & McGovern 2007; Jackson, Dugmore & Riede 2017; Hambrecht 2015; McAnany & Yoffee 2010; Turner & Sabloff 2012). Given the explanatory benefit of downscaling human-climate models, continuing to improve methods to map practical connections between people and climate at scales that more closely reflect human experience is essential (Contreras et al. 2018). In this study we present a stochastic downscaling approach that translates paleoclimate reconstructions to localized agricultural production using survey and experimental farming data. This approach is illustrated through a case study from the Central Mesa Verde region of present-day southwestern Colorado and southeastern Utah.
2. The Central Mesa Verde Context
For centuries, ancestral Pueblo farmers in the Central Mesa Verde region relied on a diverse food web to meet their subsistence needs (Cowan et al. 2012; Crabtree, Vaughn & Crabtree 2017; Minnis 1985, 2021; Rawlings & Driver 2010). Among the most important components of ancestral Pueblo diet was maize, which in some cases, made up as much as 70–80% of the ancestral Pueblo diet (Bocinsky & Varien 2017; Coltrain & Janetski 2013; Coltrain, Janetsky & Carlyle 2007; Matson 2016). Maize was not only central to ancestral Pueblo subsistence, but was also metaphorically, ritually, and ceremonially woven into their life (Cushing 1920; Ford 1994; Geib & Heitman 2015; Hegmon 1996; Wall & Masayesva 2004; Washburn 2012; Washburn & Fast 2018; Whiting 1939). Because of its importance, climate-oriented archaeology in Central Mesa Verde has focused on maize agriculture as a critical link between climate, stress, and social change.
Early research suggested that climate stress threatened subsistence and became a key driver of major demographic shifts in the region (Euler et al. 1979). Yet, community-scale studies, which compare community settlement to the size of areas that climatically support rain-fed maize agriculture (i.e., the maize niche), have shown that human-climate relationships were more complicated (Glowacki 2015; Kohler, Ellyson & Bocinsky 2020; Kohler et al. 2008; Kohler & Varien 2012; Nelson et al. 2016; Schwindt et al. 2016; Turner et al. 2003; Van West & Dean 2000). To build on these studies we must understand why communities reacted differently to climate stress, which requires continuing to refine how we localize our models. For example, the most robust models of human-climate dynamics in the Central Mesa Verde have been produced using complex models that aggregate disparate climate proxies and settlement data and project them down to the community scale (e.g., Bocinsky et al. 2016; Kohler & Varien 2012). These models were essential for complicating interpretations of climate determinism and introducing novel techniques for approximating local patterns from regional data. However, they also run the risk of smoothing over important distinct local conditions that meaningfully impact archaeological models, such as community access to perennial springs or multi-component occupations and can frequently overestimate paleoproductivity. To minimize these issues, local inferences must be incorporated into new models to more closely examine the community-specific experiences of climate stress.
The goal of this study is to more accurately reconstruct the downstream subsistence consequences of climate at individual communities. To do this, high-resolution survey data and experimental farming data are integrated with paleoclimate estimates using a Monte Carlo approach. We illustrate the utility of this stochastic method at Far View Community (A.D. 700–1250), which is among the longest-lived settlements in Central Mesa Verde containing more than 59 residential structures including a great house complex, 68 pit structures, five towers, and a possible biwall (Figure 1). By A.D. 1000, the community reached a peak population of at least three hundred residents. The Far View community has also been the subject of more than a century of archaeological research, including a recent full-coverage survey (Glowacki & Field 2024). Given the longevity of the community and the wealth of available landscape-scale archaeological data, Far View is an ideal case study for integrating paleoclimate and social data at local scales.
Figure 1
The Far View Community.
3 The Maize Productivity Model
To begin, annual maize productivity at Far View is estimated using four steps: 1) defining the local catchment area; 2) identifying potential farming areas in the catchment; 3) determining which of these farming areas were arable by incorporating climate constraints, and 4) estimating productivity yields in the arable areas. We then compare maize productivity with community needs to estimate the timing and scale of maize surpluses and deficits. Finally, to model people’s ability to change subsistence practices in response to outcomes (Haldon et al. 2020), we re-calculate maize availability by accounting for surplus storage and greater-than-average yields due to additional intensification strategies. Through these steps we anchor paleoclimate reconstructions to community- and human-scale data and estimate the impact of local, long-term maize productivity on community change.
3.1 Define the Local Catchment
Previous research at Far View has included extensive excavation at several important residential structures (Fewkes 1917, 1919; Lister 1964, 1965, 1966) and survey (Rohn 1977). These data have recently been augmented by a 110-hectare full coverage survey focused on comprehensively studying settlement and agricultural practices within the community (Glowacki & Field 2024). Although the full-coverage survey provided a fine-grained record of residential and agricultural features in the Far View landscape, it is possible that some potential agricultural areas associated with the community were outside of the survey boundary (Glowacki & Field 2024; Rohn 1977). Therefore, the community catchment was defined by incorporating outcomes of the full coverage survey, which accounted for the spread of all archaeological evidence and topographic features associated with the community (e.g., nearby drainages and adjacent mesa top areas), with the more conventional use of a 2-cost km boundary.1 Integrating both measures resulted in a boundary that was precisely defined while also roughly equivalent to the mean size of community catchments on the Mesa Verde cuesta between A.D. 725–1260 (mean = 3.16 km2; Reese, Glowacki & Kohler 2019, Supplemental Table 1). As a result, the community catchment was defined as a ~330 ha (3.3 km2) area on the northern extent of Chapin Mesa (Figure 2).
Figure 2
Demonstration of Steps 1–2 of the productivity model.
3.2 Identify Potential Farming Areas in Local Catchment
Survey, excavation, and remote sensing data show that Far View residents farmed two types of areas: 1) large contiguous mesa top areas with gentle, southward facing slopes and deep soils, and 2) narrower areas along drainages that were enhanced with check dams — small stone walls built perpendicular to drainages — and terraces to reduce erosion, slow run-off, and capture moisture and silt. Due to the southern aspect and relatively deep soil, mesa tops were optimal for non-intensive farming and required fewer landscape modifications than drainages (although there were also some check dams on the mesa top). Mesa top and drainage farming may have also resulted in different average yields (Ermigiotti et al. 2023). Because of the differing physiography, labor requirements, and potential yields, maize productivity on the mesa top and drainage farming areas was estimated separately (Figure 2).
Potential mesa top farming areas are large, contiguous areas (≥100 m2) with a slope less than ~6 degrees that were not used for other activities (e.g., residences).2 Over time, areas used for residential activities increased, thereby shrinking potential mesa top farming locations. For example, during peak population, potential mesa top farming areas equaled about 80 ha; however, during periods when population was less than 100, these areas equaled about 115 ha. Potential drainage farming areas are in or near drainages where intensification features, like check dams, intersect with natural run-off channels. Check dams identified during survey were relatively small (on average about 4 m long), closely spaced in active drainages (~9 m between check dams in the same drainage), and usually on slopes between 6 and 17 degrees (Glowacki & Field 2024). Unfortunately, many check dams that may have once existed were likely disturbed or washed away (Glowacki & Field 2024; Ives et al. 2002; Rohn 1963). Consequently, a predictive model based on recorded drainage locations and known characteristics of check dams (n = 164) was used to project the location of potential additional check dams across the Far View landscape (Figure 2). The predictive model determines check dam locations by: 1) using the SAGA Channel Network tool in QGIS (Conrad 2003) to identify channel networks from a 1 m LiDAR-derived digital elevation model (DEM, U.S. Geological Survey 2023); 2) locating where channel networks intersect with areas resembling known check dams (i.e., areas ≥10 m2 and containing a slope between 6–17 degrees); and 3) plotting check dams in points of intersection that are at least 9 m apart. The volume of channel networks and potential farming locations may increase in years with higher precipitation. To allow for increased drainage production in higher precipitation years without dramatically increasing the number of variables, three drainage farming scenarios representing high, moderate, or low precipitation conditions were created. The high drainage farming scenario contains 4.2 ha of potential farming land and 1,401 check dams, the moderate scenario contains 1.2 ha of potential farming land and 398 check dams, and the low scenario contains 0.2 ha and 61 check dams.
3.3 Identify Arable Areas considering Climate Constraints
Maize requires at least 30 cm of precipitation and 1,800 growing-season degree days in Central Mesa Verde, thus variation in temperature and precipitation affects yield rates profoundly (Bellorado 2011; Benson 2011; Bocinsky & Kohler 2014; Ermigiotti et al. 2023; Shaw 1988). Therefore, paleoclimate reconstructions were used to identify which potential farming areas were likely arable (Bocinsky et al. 2016). Climate constraints were incorporated into the model in two ways (Figure 3). First, for each year, one of the three potential drainage farming scenarios was selected based on annual precipitation amounts (Bocinsky et al. 2016). Years with reconstructed annual precipitation above 50 cm are assigned the high drainage scenario, years with 38–50 cm of precipitation had the moderate scenario, and years with less than 38 cm of precipitation had the low scenario. The selected drainage and mesa top farming areas were then combined and clipped using a maize niche model called Paleocar (Bocinsky et al. 2016)3 that identifies the locations meeting the minimum climate conditions necessary for maize growth (Figure 2). More than three-quarters of the catchment was within the maize farming niche during the occupation of the community. Therefore, years with 80 ha of potential farming area likely realized 60 ha of arable farming area given the constraints of the maize niche. At times, the constraints of the maize niche were profound, resulting in years when there were no arable farming areas.
Figure 3
Demonstration of Steps 3–4 of the productivity model.
3.4 Estimate Productivity Yields in Arable Areas using Experimental Farming Data
Although total crop failures in places like Far View were probably rare (Doolittle 2000), experimental farming data shows that, in certain instances, not all arable plots produce a crop (Ermigiotti et al. 2023). This variation relates to differences in local environmental conditions, soil care, planting time, plant density, pollinating strategies, seed selection, and irrigation (Bellorado 2007; Bocinsky & Varien 2017; Ermigiotti et al. 2023; Weber & Seaman 1985; Whiting 1939), most of which are archaeologically unknowable. To account for these conditions, we stochastically sample experimental farming data, refining the estimate of the percentage of arable mesa top and drainage farming areas that produced a crop, and the total yields from productive areas. For each iteration, random productivity rates were sampled and applied to the arable mesa top and drainage farming areas (Figure 2). Potential productivity rates were based on experimental data showing that 75–85% of the plots within the maize niche produced a crop (Bocinsky & Varien 2017; Ermigiotti et al. 2023). Yields were then calculated by multiplying the productive arable farming area (ha) by a randomly selected yield amount (kg/ha). Yield amounts were derived from experimental farming data, which show considerable variability among productive gardens (1–5,602 kg/ha, Ermigiotti et al. 2023). Notably, gardens associated with ancestral check dams had less variability and a higher mean productivity (13–1,4681 kg/ha) than other gardens (Ermigiotti et al. 2023). Thus, under average climate conditions, drainage areas should produce the most maize per hectare, whereas mesa top areas should produce the most maize per ha under best-case conditions. These differences likely mean that farmers dedicated most of their labor to mesa-top farming, hoping to amplify production in areas where they could get the greatest returns, yet still devoted some effort to drainage farming, where returns were more constant and less dependent.
The first four steps (section 3.1–3.4) were integrated into a Monte Carlo simulation that was iterated 100 times to produce an average estimate of maize productivity during Far View’s occupation. Next, the following steps were used to compare maize productivity with community needs and to account for the impact of maize surpluses and deficits on farming efforts in the community.
3.5 Compare Productivity to Maize Needs
People did not only use annual maize yields for subsistence. Some portions were needed to seed next year’s crop, feed animals, or were lost due to spoilage (Kuijt 2017; McCaffery, Miller & Tykot 2021; Rawlings & Driver 2010; Schurr, McLeester & Countryman 2021). Ethnographic and archaeological data suggest the percentage of yields withheld for seed typically range from 10–35% (Kuijt 2017; Schurr, McLeester & Countryman 2021; Van West 1994; Whiting 1939). Although seed withholding in the U.S. Southwest may have been as low as ~10% (Pool 2013; Van West 1994; Williams 1989), maize was also withheld to feed domesticated turkeys, which were important to Pueblo subsistence (Crabtree, Vaughn & Crabtree 2017; McCaffery, Miller & Tykot 2021; Rawlings & Driver 2010). Therefore, to estimate maize yields available for subsistence at Far View, 25% of annual maize yields were removed for planting or non- human consumption. The remaining value (in kg) was multiplied by 3,500 (i.e., the average calories in one kg of maize).
Yields available for consumption were then compared to the needs of the community, which was calculated by multiplying annual per capita needs by the number of residents each year. Individual caloric needs were calculated using body size and caloric intake estimates of ancestral Pueblo people (Matson 2016; Palkovich 1980). The calculated estimates suggest that each Far View resident required between 746,000–817,000 calories annually, of which 520,000 would be acquired through maize.4 Community-wide demographic estimates were taken from Field & Glowacki (2024:492), who assigned temporal occupations to residential structures by comparing diagnostic ceramics to a seriation of calibrated ceramic data from the Village Ecodynamics Project (VEP 2015). The number of occupants for each structure was estimated by averaging population based on the number of associated kivas and the volume of architectural rubble (Duwe et al. 2016; Field & Glowacki 2024). The comparison of maize availability and needs produced a rolling estimate of annual maize outcomes; surpluses were defined as any year when more maize was produced than needed and deficits were defined as any year with less than needed.
3.6 Enabling use of Surpluses
When farmers produced a surplus, we assumed the entire surplus would be stored to offset or minimize potential near-future shortages when available (Whiting 1939). However, food stores were increasingly susceptible to spoilage over time, which even in arid or semi-arid environments could reach 40% per year (Abass et al. 2014; Adams 1976; Bradfield 1971; Ford 1968; Hough 1915; Kintigh & Ingram 2018; Kuijt 2015, 2017; Maziku 2019; Ngowi and Selejio 2019; Shee et al. 2019; Suleiman, Rosentrater & Bern 2015; Titiev 1944; Whiting 1939). Thus, for this study, maize surpluses were made available for subsistence for three years with unconsumed surpluses shrinking by 10% after one year of storage due to spoilage, 35% after two years of storage, 70% after the third year, and were unusable after four years. Consequently, the amount of maize available for subsistence needs each year was estimated as a rolling total that accounted for yearly production, subsistence needs, leftover surpluses, and ongoing spoilage.
3.7 Enabling Intensification to Account for Deficits
Faced with maize deficits, Far View residents would have pursued strategies that included food diversification, trade, and intensification (Crabtree 2015; Crabtree, Vaughn & Crabtree 2017; Hegmon 1996; Kohler 2010). However, climatic downturns that significantly affected maize productivity at Far View likely impacted the access to food across the region (Cook et al. 2004; Van West & Dean 2000), reducing the potential for Far View residents to diversify or trade their way out of a deficit. Considering this, the most reliable strategy for mitigating maize deficits at Far View was intensifying local production. Archaeological evidence of a network of diversion ditches, water retention features, and check dams show residents did intensify their local catchment to meet subsistence needs, although it is unclear how much these efforts increased maize production (Glowacki et al. In prep.; Glowacki & Field 2024; Rohn 1963, 1977; Stewart 1940; Stewart & Donnelly 1943). Therefore, to account for the ways that farmers would have increased their intensification strategies to address the threat of maize shortages, we estimate the deficits that could have been avoided by a 1–20% increase in the amount of productive land and average yield per year. We use a threshold of 20% because we expect that increases at-or-below 20% would have been feasible for farmers in Far View. For example, intensifying by 20% during peak population would require each farmer to tend to 0.1 ha more farmland, produce 100 kg more maize on that land, and spend about 8 more days a year laboring (Asnah et al. 2018; Pimentel 2009; Twomlow et al. 1999).5
4. Illustrating Model Performance
To aid reproducibility and understanding of how the model functions, we describe the outputs from a single decade (A.D. 990–999). Between A.D. 990–999, nearly all the Far View catchment was within the maize growing niche, with the notable exception of A.D. 990 (Figure 4). On average, there were 66 ha of potential mesa top farming areas, 2.2 ha of potential drainage farming areas, and more than 38 cm of annual precipitation during this period. Due to the generally good conditions and sizable farming areas between A.D. 990–999, maize productivity in the community averaged 38,440 kg; however, variations in maize niche size and precipitation caused lower production in certain years. In A.D. 990, high precipitation resulted in potentially large drainage farming areas. Yet, temperatures were also colder that year, which erased any benefits of higher-than-normal precipitation by limiting the maize niche size overall. Consequently, in A.D. 990, the amount of productive farmland in the community was only 10.6 ha and productivity was very low. In A.D. 995, lower-than-normal precipitation limited the size of potential drainage farming areas. At the same time, however, more than 80% of the community was in the maize niche. As a result, the relatively large arable mesa top farming areas counteracted the low drainage production and enabled high maize productivity during that year.
Figure 4
Annual maize productivity results between A.D. 990–999 (each dot is a single iteration, and cross bars show the mean). (A) Proportion of FV catchment in maize niche each year. (B) Amount of precipitation (cm) in the catchment. (C) Distribution of potential mesa top and drainage farming areas used for maize production. (D) Distribution of yields in arable mesa top and drainage farming areas and total simulated maize productivity.
Despite good conditions, maize production threatened to fall below needs throughout the decade. Without additional intensification, every year except A.D. 996 would have resulted in a maize deficit, even when accounting for surpluses that had been intermittently produced from A.D. 887–990. Fortunately, maize deficits could have been alleviated with anywhere from an 8–34% increase in the size of productive farmlands and yields. Cumulatively, results from the decade show how life at Far View depended on the size of the maize niche, the success of mesa top farming, and intensification. Even though intensification strategies were most likely focused on the mesa top, which consistently provide the best returns, augmentation of drainage areas would have been a key for supplying marginal increases in productivity.
5. Social Consequences of Maize Production
From A.D. 700–1300, average maize production was about 40,000 kg a year, grown on an average of 66 ha of the mesa top and in 2 ha of drainages (Figure 5). Yields accounted for around 140 million calories annually, nearly double the average needs of the community but less than peak needs (Figure 5). Without increased intensification, we estimate 129 years when the community would have experienced deficits (Figure 6). These shortages varied in severity and duration but were most significant between A.D. 975 and 1100, when population peaked. These deficits most often occurred because of years when none of the catchment was within the maize niche (n = 34).6 However, low-to-moderate degrees of intensification could have avoided at least 96 of these deficit years. The most persistent deficits (n = 33), or deficits that would have occurred despite more than a 20% increase in productivity, took place almost exclusively between A.D. 1000–1100 (Figure 6).
Figure 5
Maize productivity results for entire study period: (A) Annual and ten-year average of arable mesa top farming areas; (B) Annual and ten-year average of arable drainage farming area; and (C) Annual and ten-year average of total maize productivity (calories) in the community.
Figure 6
(A) Annual and ten-year average of total maize productivity (calories) and annual maize needs (calories) derived from smoothed population estimate; (B) Maize accumulation (calories), showing years with a maize deficit in red and years with a maize surplus in green; (C) Years with maize deficit; and (D) Years with persistent deficit.
Outcomes from this model, particularly the timing and severity of persistent deficits in the 1000s, help explain the decline of the community in ways that are not clear in regionally derived reconstructions. For example, regionally derived reconstructions show that the size of the Far View community decreased when the maize niche size bottomed out (early 1200s; Bocinsky et al. 2016; Schwindt et al. 2016; Figure 7), but in other periods of small maize niche size (e.g., the 1090s) population increased. In this case, addressing why the Far View population would decline during one period but not during others is necessary. To answer this question requires more than reconstructing maize niche size, but rather reconstructing probable maize yields and community demands. Our model is built to produce the latter, and results show that only under conditions of high demand did small maize niche size result in yields that could not meet community needs. Our model suggests that population thresholds were a key driver in maize vulnerability during the eleventh century, which set the stage for how the community responded to challenges faced in the twelfth and thirteenth centuries. To illustrate the explanatory power of our approach, we review results within the context of three periods that contain distinct patterns of demographic change, maize productivity and subsistence success/stress, and likely farming strategies.
Figure 7
Estimate of community size and maize niche size produced from regionally derived reconstructions (Bocinsky et al. 2016; Schwindt et al. 2016).
5.1 Growth and Subsistence Success (A.D. 700–980)
During the formation of the community and in the following two-plus centuries of growth (A.D. 700–900), subsistence needs were consistently met. During this time, the population never grew to more than around 150 people and mesa top farming was likely emphasized. During the ninth century, productive capacity may have peaked due to relatively high-quality climate conditions and consistently high percentages of the catchment within the maize niche (Figure 5). While farming conditions were slightly reduced from A.D. 900–980, farmers continued to achieve near-constant success in maize production, allowing them to overcome short-term maize shortages at the end of the tenth century.
5.2 The Height of the Community and Vulnerability (A.D. 980–1100)
Population grew rapidly in the late tenth century and consequently, subsistence pressures mounted. Between A.D. 980–1000, intensification strategies that increased productivity on the mesa top, including water storage and rain-fed irrigation, as well as strategies that opened other parts of the landscape to cultivation, such as check dams, became essential to subsistence success (Glowacki et al. In prep.). That the model indicates check dams were necessary by the late tenth century supports prior inferences that check dams were used before A.D. 1000 (Reese 2020; Rohn 1977). Intensification during the eleventh century would have required additional labor, which could be supplied by the growing population during that time. However, intensification strategies also required social cohesion, inter-household trust, and cooperation, which would be harder to maintain in a community of 300 than a community of 100 (Alberti 2014; Bernardini 1996; Johnson 1982).
Between A.D. 980 and 1100, population peaked and the potential for subsistence deficits grew, even if people adopted critical mitigation strategies. The eleventh century marked the height of subsistence vulnerability at Far View. The number of people to feed had reached its peak, climate had become marginally worse compared to prior conditions, and persistent deficits became common, reversing the long-term subsistence successes that had substantiated the growth of the community (Figures 4, 5, and 7). To contend with these conditions water storage, irrigation, and check dam strategies were likely amplified to peak capacity, perhaps straining the practical limit of the catchment and social networks within the community. Therefore, at the time when subsistence vulnerability in the community peaked, it is possible that some of the social and intensification strategies that once made the community resilient had begun to erode.
5.3 Consequences of Subsistence Vulnerability and Community Decline (A.D. 1100–1250)
After more than a century of ongoing subsistence pressure, about a third of the residents left the community around A.D. 1100. The reduced size of the community enabled remaining residents to regain some subsistence security for at least a few generations (A.D. 1100–1180). Despite the rebound in subsistence success, the population at Far View would not grow again. Around A.D. 1180, population declined precipitously and by the mid-thirteenth century people had left entirely. The decline of Far View was influenced by many interacting and dependent processes. In the prior period, Far View residents had likely become dependent on relatively complex, labor-intensive subsistence strategies. While these had the potential to increase maize productivity, they also demanded a large labor pool and coordinated maintenance and operation. Therefore, a shrinking community population had counteracting effects that amplified scalar stressors at Far View (Alberti 2014; Bernardini 1996; Johnson 1982); fewer people reduced subsistence needs while simultaneously reducing capacity to operate or manage pre-existing subsistence strategies, which increased the vulnerability of community subsistence. Just as residents were beginning to regain some subsistence stability, a severe regional drought (i.e., the so-called megadrought from A.D. 1130–1150) and two periods of acute climate stress — marked by poor, unpredictable climate conditions — took place (Figure 8; Cook et al. 2004; Field & Glowacki 2024; Van West & Dean 2000). Yet, within two generations, the worst climate conditions at Far View occurred, resulting in the first part of the 1200s having the lowest maize productivity in community history. While these climate conditions did not cause long-term maize shortages because the maize needs were comparably low, they were acute shocks to the subsistence efforts of remaining Far View residents. Additionally, the long-occurring bimodal precipitation pattern that brought relatively heavy precipitation to Central Mesa Verde during summer and winter months was disrupted in the mid-thirteenth century (Cordell et al. 2007; Van West & Dean 2000). The combined effect of short-term climate stressors, a disruption of long-term climate patterns, and a decreased ability to effectively operationalize intense subsistence strategies meant continued subsistence struggles and ultimately, a growing set of reasons to leave the community.
Figure 8
(A) Far View community population; (B) Persistent deficit years; (C) Low precipitation years (<35 cm); and (D) Periods of acute climate stress (see Field & Glowacki 2024).
6 Conclusion
The results from our model suggest that the growth of Far View was enabled through long-term maize success in the 8th through 10th centuries. However, persistent maize shortages — marked by increasing needs and insufficient productivity — became common in the 1000s. Then, in the 1100s, any potential rebound that may have been possible due to decreased demands was thwarted by low precipitation and acute climate stress, which ultimately put the community in a poor position to cope with the shocks of the early 1200s. Our localized model, which highlights the agricultural challenges of the 1000s and how those challenges interact with climate stress in later centuries, presents a more nuanced history of the growth and decline of the Far View community. Ultimately, the results of our model enabled us to address why settlement dynamics changed at Far View in relation to climate stress.
Our case study demonstrates the explanatory benefit of downscaling human-climate models to local conditions, where the downstream subsistence consequences of climate stress play out. Here, paleoclimate reconstructions were downscaled to a meaningful measure of agricultural productivity using survey data and a stochastic sampling of experimental farming data. Although the specific survey and experimental data used in this study will not be relevant to many other contexts, we expect our methods could be a useful road map for others due to the translatability of stochastic frameworks as a means of grounding archaeological models in human dynamics (Wolpert et al. 2024). The essential elements to this approach involve: 1) using survey data and paleoclimate reconstructions to define productive agricultural areas within a catchment more accurately; 2) stochastically sampling experimental farming data to translate agricultural areas to yields; and 3) comparing yields to community needs (population) that are also grounded in local survey data. Using this approach archaeologists studying human-climate dynamics will be able to more closely study the complex, diachronic processes that affect people’s resilience to climate stress.
Data Accessibility Statement
All code needed to implement the method presented in this study is available at https://github.com/sfield2/Link-Paleoclimate-Survey-FarmingData. Due to the sensitive nature of certain data, not all data necessary to replicate the analyses presented in this research are accessible at the above GitHub repository; however, all unique functions and code needed to integrate paleoclimate, experimental farming data, and survey data to estimate maize production are available and demonstrated through publicly accessible data within the repository. Additionally, all tabular outputs from our analyses and code needed to recreate figures and maize deficits are also available in the repository.
Notes
[1] The 2-cost km catchment encompasses areas that can be accessed from the center of the community in an identical amount of time as if someone walked 2 km on an established, level walking path, and is based on experimentally and ethnographically informed research that measures the frequency of household movement via the accumulation of household materials (Reese, Glowacki & Kohler 2019; Varien 1999).
[2] Most check dams in FVC were found in slopes between 6–18 degrees. Therefore, mesa top farming areas were defined using 6 degrees as an upper slope boundary.
[3] We use the rasterized temperature and precipitation reconstructions developed by Bocinsky et al. (2016). Following Bocinsky et al. (2016), areas within the maize niche are any location with precipitation greater than or equal to 350 mm and annual growing degree days greater than or equal to 1800. For other configurations and developments of the PaleoCAR function see Reese (2020).
[4] Annual maize needs are based on the assumption that Far View residents relied on maize at the same rate documented in other ancestral Pueblo communities (i.e., 70%, Bocinsky & Kohler 2014; Matson 2016).
[5] This is assuming half of the population was responsible for farming any given year and producing a maize crop that required at least 700 hrs/ha annually.
[6] Years when the entire catchment was outside the maize niche suggest that rain-fed maize agriculture was not possible at those times. However, irrigation or intensification could have made productivity possible, and consequently it is unlikely that there was no productivity even in years when the catchment was outside the niche.
Acknowledgements
This research would not have been possible without the assistance of all those who contributed to the Far View Archaeological Project (FVAP, Glowacki and Field 2024). We also thank the National Park Service for their collaboration and use of data related to Far View. Thank you to Kyle Bocinsky and Kelsey Reese who provided early insights regarding paleoclimate reconstructions and additional thanks to Carrie Heitman, Mark Schurr, and Ian Kuijt for their thoughtful contributions and recommendations. Thank you as well to the two anonymous reviewers who provided important insights that were used to increase the quality and clarity of this research.
Competing Interests
The authors have no competing interests to declare.