CHAPTER 3: CONCEPTUAL MODELS OF VULNERABILITY
TO DRY PERIODS
Return to Table of Contents
My approach to advancing understanding of conditions
that affect human vulnerability to dry periods is to evaluate four prominent and often implicit conceptual models used by
archaeologists and others investigating past human adaptations in dry
climates. These models are used to
explain and predict spatial and temporal differences in vulnerability to dry
periods and they each emphasize different demographic and environmental conditions
assumed to influence this vulnerability.
Neither I nor the researchers that have employed the models I evaluate assume that demographic and environmental conditions are the only conditions that contribute to vulnerability to dry periods. These conditions and changes in these conditions, however, are often emphasized (individually or in combination) and form an important link in the chain of causal argument regarding climatic influences on human behavior. The individual contribution of a specific condition to dry-period vulnerability has been argued because in the context of assumed resource marginality, changes in either resource demands (represented by demographic conditions) and/or changes in resource supplies (represented by environmental conditions) may have created population-resource imbalances that increased vulnerability to dry periods. The indicators I use to represent differences in these conditions (settlement population levels, watershed population density, settlement locations relative to perennial rivers, and average annual precipitation levels) are also only a subset of characteristics that may influence resource demand and supply.
I identified these models by considering arguments, expectations, and assumptions regarding climatic influences on human behavior prevalent in U.S. Southwestern archaeological studies and in modern studies of vulnerability to natural hazards conducted by researchers in other disciplines. This is the first archaeological effort in the U.S. Southwest or elsewhere, to my knowledge, to explicitly identify these models and systematically examine over an extended period the influence of demographic and environmental conditions on vulnerability to dry periods in order to test the validity of these models. It is important to evaluate these models because they lack empirical scrutiny, are seldom identified explicitly but are rather embedded in other arguments, and rely on an unverified assumption of widespread resource marginality due to low precipitation in the region.
In this chapter, I describe these models and provide examples of their use and discuss how they have affected interpretations of vulnerability to dry periods as well as important events in the prehistory of the U.S. Southwest. The models are summarized in Table 3. I also provide an initial description of how I evaluate each model with greater detail provided in Chapter Five.
See Table 3.1.
Aridity Model of Vulnerability to Dry Periods
Aridity models emphasize the contribution of low precipitation and resource marginality to vulnerability to dry periods. Vulnerability in this model is understood in terms of the biophysical conditions of the environment (resource marginality and aridity), ignoring characteristics of human populations (Liverman 1990b). Biophysical vulnerability "is a function of the frequency and severity (or probability of occurrence) of a given type of hazard" (Brooks 2003:4) such as a dry period. Social, demographic, and other conditions of human populations exposed to dry periods and the contribution of these conditions to their vulnerability to dry periods are either ignored, not emphasized, or assumed to be small in biophysical models of vulnerability to natural hazards.
An emphasis on the contribution of biophysical conditions to vulnerability to climatic hazards is widely applied by engineers and economists in the technical literature on disasters (Fussel 2007) and within global environmental change studies (Liverman 1990b:30). An emphasis on biophysical conditions has its theoretical origins in the risk-hazard vulnerability approach originating in geography. Two examples of this approach are vulnerability to sea level rises and vulnerability to floods. Topographic contours are used almost exclusively to identify the extent of vulnerability to these hazards (e.g., Titus and Richman 2001). Regarding vulnerability to global environmental change, a biophysical approach "implies that in order to understand and delimit vulnerability, we just need to know how and where the physical environment will change. Physical indicators will then provide adequate insight into the populations at risk" (Liverman 1990b:30).
Archaeological acceptance of a biophysical model of vulnerability to dry periods (and other climatic conditions) is pervasive in the U.S. Southwest because of the assumption of resource marginality and widespread vulnerability to relatively low precipitation in the region (as I discuss in the "Resource Marginality” section in Chapter Two). The assumption of resource marginality when employed to explain social and cultural changes in the U.S. Southwest essentially draws the contours of vulnerability to dry periods around the entire region due to relatively low and variable precipitation. As a result, studies of human-environment interactions and climatic influences on human behavior in the U.S. Southwest are usually studies of risk and the contribution of marginality and climatic event severity to this risk (e.g., Doyel and Dean 2006; Graybill et al. 2006; Gumerman 1988; Larson et al. 1996; Minnis 1985; Nials et al. 1989; Rautman 1993; Tainter and Tainter 1996).
Studies that emphasize quantitative estimates of precipitation levels relative to "normal" or average conditions and compare severe conditions to coincident behavioral changes signal the use of an aridity model. For example, Stahle et al. (2007) and Cook et al. (2007) use tree-ring reconstructed summer Palmer Drought Severity Indices (PDSI) to identify decadal droughts more severe and prolonged than any witnessed during the modern instrumental period. Cook et al. (2007; and references contained therein) identify temporal correlations between severe dry periods and a number of important socio-cultural events including the 14th century decline of complex Mississippian chiefdoms and famine, disease, and village abandonment in the Puebloan region of New Mexico in the 16th century. Stahle et al. (2007:136-140) also use the temporal coincidence between identified droughts and historical information on the environmental conditions and activities of Euroamerican and Native American societies throughout the U.S. to validate the PDSI reconstructions. Although Stahle et al. (2007:136) find a number of temporal correlations between severe dry periods and social and environmental events, they also find a number of "disagreements" between these dry periods and known historical events. Stahle et al. (2007:136) attribute these disagreements to climate proxy reconstruction errors or the fact that “tree-ring data integrate moisture conditions during and sometimes preceding the growing season, and may not well represent fall or winter conditions.” I suggest that these disagreements between dry periods and known historical events also strongly suggest variation in human vulnerability to dry periods driven by other factors in addition to weaknesses in the climate reconstruction's ability to accurately identify relevant precipitation conditions and their severity.
Explanations of regional-scale abandonments that attribute them to catastrophic climate events, such as dry periods, treat vulnerability as a regional-scale biophysical condition and assume endemic shortfalls and widespread vulnerability due to low precipitation. This assumption of widespread vulnerability ignores or over-rides any intra-regional social and environmental diversity that influences vulnerabilities. For example, abandonment of the San Juan Basin in the mid-1100s and the Mesa Verde region in the late 1200s has been (Douglass 1929; Judge 1989, and others) and continues to be (Benson et al. 2007) attributed to severe dry periods. Differences in vulnerability across these regions due to varying demographic and environmental conditions and how these differences might have affected dry-period impacts and responses are often not systematically considered. Van West’s (1994) spatially detailed study of soil types and modern crop yields and potential variation in impacts of the late 1200s drought on agricultural productivity is an early and notable exception (see also recent work by Axtell et al. 2002, Kohler et al. 2007, and Schollmeyer 2009). In central Arizona, dry periods combined with catastrophic flooding and associated stream channel changes along the Lower Salt River in the late 1300s have been hypothesized as the cause of the depopulation of settlements irrigating from the Lower Salt River (Graybill et al. 2006, Nials et al. 1989; Gregory 1991). Again, differences in vulnerability based on settlement locations (e.g., near or distant from the floodplain) and settlement-scale demographic conditions that affected the demand for resources are typically not considered.
Dry-period influences on regional abandonments have been argued for many places and times (e.g., Weiss and Bradley 2001). Examples include the Maya of Mesoamerica ca. A.D. 1200 (Gill 2000; Hodell et al. 1995), Cahokia and associated settlements of the American Bottom and Mississippi River Valley ca. A.D. 1100 to 1245 (Benson et al. 2009), the Tiwanaku of Bolivia-Peru ca. A.D. 1100 (Binford et al. 1997), and the Akkadians of Mesopotamia ca. 4200 B.P. (Weiss et al. 1993), to name only a few. Dry periods are seldom considered the only factor responsible for these regional depopulations; however, variability in vulnerability across a region is not addressed in aridity models that emphasize widespread vulnerability to dry periods due to low average precipitation conditions.
To assess the utility of an aridity model for explaining vulnerability to dry periods, I examine the relationship between dry-period severity and residential abandonment throughout central Arizona from 1200 to 1450. I use the relationship between dry-period severity and residential abandonment as an indicator of the extent of vulnerability to dry periods within six watersheds with a long-term history of occupation. By geographically disaggregating the study area into watersheds, I test the assumption of vulnerability to dry periods as a regional-scale, biophysical condition. The aridity model and the assumption of regional-scale resource marginality would be supported by strong and sensitive relationships between dry-period severity and residential abandonment in all watersheds.
Demand Models of Vulnerability to Dry Periods
Demand models emphasize the contribution of demographic conditions to vulnerability to dry periods. The demographic conditions I consider are settlement population levels and watershed population density. These conditions define, in part, resource demands, the rate of consumption of resources, and the extent of labor available to invest in strategies to manage shortfalls. Settlement population levels reflect the cumulative decisions of people to stay or leave a locale--a settlement or watershed area--and to allow others to move into that locale. Exploration of the role of population levels and density is a window into the role of human decisions in creating or ameliorating vulnerabilities to dry periods.
The effect of settlement population levels and areal population density on vulnerability to dry periods is complex and has been argued to both increase and decrease vulnerability (Meyer et al. 1998:241). For example, as population levels increase more resources are consumed and increases in production may not be able to keep pace (Malthus 2001 [1798]) especially during dry-period declines in productivity. Relatively high population density in an area can also increase vulnerability by limiting mobility as a strategy to manage shortfall risks (Binford 1983; Cordell 2000:183; Dean et al. 1994:85; Minnis 1996; Powell 1988). Population density can limit mobility if productive locations are already claimed, occupied, or hostilities restricted movement. Larger populations might also increase the tilling of relatively less productive plots of land to increase production (B. Nelson et al. 1994:61-62) and these plots could quickly become unproductive during dry periods. Thus, "as population density increases, more individuals are forced into areas of greater risk" (Reycraft and Bawden 2000:2). Larger populations can, however, intensify production and invest in infrastructure to increase resource supplies (Boserup 1965) and develop social strategies not as readily available to smaller populations, such as centralized leadership and alliances (McGuire and Saitta 1996; Wilcox et al. 2001a; 2001b), to manage shortfall risks.
Demand models are essentially 'population pressure' models wherein increasing population levels are considered responsible for some stress on humans or the environment (e.g., Smith 1972). Population pressure arguments assert that rising population levels at some point breach a threshold and behavioral responses become necessary. Where resources are assumed to be marginal, this threshold is expected to be easily and frequently breached. For example, Larson et al. (1996:219) assert "With the shift to primary reliance on domesticated crops, population levels and densities increased dramatically between A.D. 900 and 1100 and the Anasazi [northern Southwest] became increasingly vulnerable to climatic variability and extremes." Larson et al. (1996:218) describe this increase in vulnerability in the context of an "extremely marginal environment for prehistoric hunting and gathering and agricultural pursuits." Similarly, Plog et al. (1988:261) state "It is difficult to imagine environmental deterioration of sufficient severity to stimulate a long-distance move in the absence of high populations." High population levels, then, in these examples are a critical factor in understanding vulnerability to dry-period risks of shortfall.
It is important to empirically evaluate this model because expectations of the contribution of demographic conditions to vulnerability to dry periods (and other hazards) are often conflicting (as discussed above) or too ambiguous to be useful or convincing. For example, in a list of possible determinants of vulnerability to global environmental change (which includes increases in dry-period severity associated with global-scale climate change), Liverman (1990b:33) asserts "High population densities, population growth rates, and pressure on limited food, land, and water resources can make regions very vulnerable to global change." Similarly, Reycraft and Bawden (2000:2 citing work by White 1974 and Burton et al. 1978) state that "the greater the population density of a given area, the greater the damage potential of a given extreme event. Also, as population density increases, more individuals are forced into areas of greater risk." And, Cordell (2000:182), referring to upland areas of the northern Southwest, states "Population 'packing' [makes it] impossible to implement an agricultural strategy that depends on relocating fields and dwellings" thereby linking increases in population densities with agricultural vulnerability. These expectations and assertions though sometimes vague are logically appealing and amenable to empirical testing. Minimally, we need to understand at what spatial scales the suggested relationships might apply.
Investigations of the potential influence of increasing population levels on vulnerability to dry periods without considering the potential resource productivity of an area are unrealistic. Coupled with the assumption of region-wide resource marginality, however, these studies place population levels close to a threshold where changes in either population or resources may increase the risk of shortfall. Under this model, people living in settlements or watersheds with relatively high population levels would be more vulnerable to dry periods than those living in places with lower population levels. Thus, for the central Arizona study area, if vulnerability to dry periods was influenced by settlement population levels and watershed population density, the relationship between dry-period severity and residential abandonment will be stronger and more sensitive where population levels and density are highest and weaker where population levels and density are lowest. Results will clarify the influence of these demographic conditions on vulnerability to dry periods in the study area.
Supply Models of Vulnerability to Dry Periods
Supply models emphasize the contribution of the potential productivity of settlement locations to vulnerability to dry periods. The environmental conditions I consider are settlement locations relative to perennial rivers and areas of low to high average annual precipitation levels. Access to resources, as reflected by settlement locations, is one factor that affects the ability of human systems to adapt to and cope with dry periods and climatic conditions. Settlement locations reflect the decisions of people regarding where to live; thus, exploration of the role of settlement locations is a window into the role of human decisions in creating or ameliorating vulnerabilities to the risk of shortfall.
Settlement locations adjacent to perennial riverine resources, which include the associated riparian and aquatic resources, are assumed in this study and model to offer greater potential productivity than settlements located away from perennial riverine resources. The majority of wild plants used as resources by the Hohokam "are most densely and continuously distributed along riparian corridors" (Fish and Nabhan 1991:42). Agricultural potential is also greater in riverine areas where irrigated and floodplain agriculture are possible.
Likewise, settlements in areas receiving relatively more precipitation on average are assumed to have been relatively more productive than those areas that receive less. Differences in precipitation conditions are often used to explain settlement location shifting motivated by changing climatic conditions (e.g. Dean 1988; 1996). Similar assumptions regarding proximity to rivers and precipitation levels have been used to identify "agricultural primeness" in the arid regions of Africa (Miller et al. 2002). Other factors such as the extent of arable land or the quality of soils (Sandor et al. 2007) influence productivity but are beyond the scope of this study.
The influence of potential resource supplies on vulnerability to dry periods wherein vulnerability is considered a function of riverine proximity or precipitation levels is a common sense notion supported by the strong relationship between the distribution of water and settlement locations in dry climates. People living distant from perennial rivers rely on "dry-farming" or "rain-fed" farming and are widely assumed to be among the most vulnerable and sensitive to low precipitation in dry climates (e.g., Liverman and Merideth 2002:207). For example, Harry (2005:299) compares two settlement areas in the Tucson Basin (southern Arizona): one bordering a primary river with wide expanses of arable floodplain, the other along lesser watercourses and in an upland area. Harry (2005:299) "intuitively" expects that the settlements near the primary river were not agriculturally impoverished relative to the settlements along the lesser watercourses and in the upland area. In colonial Mexico, Florescano (1980, as translated by Liverman 1990a:50) argues that "the disastrous effects of drought, as in earlier times, are concentrated in the rainfed agriculture practiced by the poorest ejidatarios and campesinos, lacking credit, irrigation, fertilizers, and improved seeds."
The notion that riverine areas were less marginal than non-riverine areas and people living near perennial rivers were less vulnerable to dry periods than those living distant from perennial rivers lacks empirical verification and there are reasons to question its validity. For example, investments in irrigation infrastructure in riverine areas will increase productivity and resource supplies thereby potentially reducing vulnerability to dry periods. However, these investments in irrigation infrastructure may increase population, reduce mobility, and thereby increase vulnerability to dry periods as well as other social and environmental conditions (Anderies 2006; Ingram 2008; Janssen and Anderies 2007; Nelson et al. 2010). In a study of the impact of irrigation on drought losses in Mexico, Yates (1981) finds little support for the expectation that irrigation has the advantage of reducing climatic hazards compared with crops dependent on erratic rainfall. Rather, he finds "only a slightly smaller deviation from [agricultural] output trend in irrigated than in rainfed areas” (Yates 1981:77 as quoted in Liverman 1990a:58).
To assess the utility of a resource supply model for explaining spatial differences in vulnerability to dry periods in central Arizona from 1200 to 1450, I examine the relationship between dry-period severity and residential abandonment among rooms near and far from perennial rivers and in areas of low, moderate, and high precipitation. If vulnerability to dry periods was strongly related to the productivity characteristics of settlement locations as identified by differences in the availability and access to water, then the relationship between dry-period severity and rooms abandoned near perennial rivers or in areas with high average annual precipitation will be less sensitive and weaker than the relationship between dry-period severity and rooms abandoned distant from perennial rivers or in areas with low average annual precipitation.
Demand and Supply Models of Vulnerability to Dry periods
Demand and supply models emphasize the contribution of population-resource imbalances (at various spatial scales) to vulnerability to dry periods (Cordell and Plog 1979; Dean 1988; Dean 1996; Larson 1996; Larson and Michaelsen 1990). Population-resource imbalances were "probably a fact of life for most prehistoric groups, both because they sometimes grew too fast and because of unforeseeable decreases in the resources available to them" (Cordell and Plog 1979:411). Emphasis on population-resource imbalances are not limited to climate-human behavior studies. Rather, "Interpretations of Southwestern cultural patterns are frequently based on the assumption of high population density, low plant and animal biomass [resource marginality], and resultant population/resource imbalances" (Powell 1988:182).
Attempts to consider both potential resource supplies and demand to infer shortfall risks pose a substantial challenge in archaeological research. It is a challenge because we have inadequate information about the plethora of variables involved in inferring both supply (wild and cultivated food resources) and demand (population levels). For example, identifying the extent of a food supply involves an understanding of the water requirements of prehistoric maize varieties, the proportions of cultivated and wild foods that comprised diets, yields of cultivated and wild foods, and many other factors (Minnis 1985:99-155 provides an extended discussion of these problems).
Despite these challenges, reasonable attempts have been made that provide approximations of periods when population-resource imbalances likely resulted in increased vulnerability to food shortfalls. One such effort is a study developed by Minnis (1985) for the Mimbres region of southwestern New Mexico focused on the A.D. 600 to 1250 period. Minnis documented potential resource supplies over time with estimates of crop success, wild food productivity, and food stress using precipitation and streamflow reconstructions. He compared these estimates to potential resource demands identified by variation in population levels. The resulting effort identified periods when population-resource imbalances and associated vulnerability to dry periods were most likely.
Examining the influence of population-resource imbalances is important because it allows us to see if these imbalances emerge as a consistent influence on vulnerability to dry periods over time. As I discussed in Chapter Two, people have a variety of strategies for managing shortfalls risks and their vulnerability to dry periods. The effectiveness or changes in the effectiveness of these strategies might be a more important factor in vulnerability to dry periods than simple formulations of the potential balance between resource supply and demand. As stated by Dean et al. (1994:86), "Many congruences [between past environmental variability and regional demographic trends] establish the importance of demographic and environmental variables as integral factors in sociocultural stability, variability, and change in this harsh and variable region. On the other hand, many failures of the archaeological record to fulfill the expectations of the models indicate that the effects of population and environmental fluctuations were mediated and transformed by sociocultural factors." Evaluating models of vulnerability to dry periods that emphasize population-resource imbalances, then, help us understand the extent to which these imbalances consistently influenced vulnerability to dry periods.
To assess the utility of this model for explaining differences in vulnerability to dry periods, I combine elements of the demand and supply models (above). I test the expectation that the extent of population-resource imbalances influenced vulnerability to dry periods by comparing the strength of relationships between dry-period severity and residential abandonment where demands were relatively high (settlements located in watersheds with high population density and settlements with high population levels) and supply low (settlements located distant from perennial rivers and in areas of low precipitation) to the relationship where demands were relatively low (settlements located in watersheds with low population density and settlements with low population levels) and supply high (near perennial rivers and in areas of high precipitation). If vulnerability to dry periods was influenced by population-resource imbalances, then the relationship between dry-period severity and residential abandonment will be stronger and more sensitive where demands were high and supplies low (population-resource imbalances most likely) than where demands were low and supplies high (population-resource imbalances least likely).
Conclusion
In this chapter, I describe four conceptual models that emphasize the influence of aridity and demographic and environmental conditions on vulnerability to dry periods. These models reflect logical relationships between the demand and supply of resources and vulnerability to dry periods in areas where resources are considered marginal and the risk of food shortfalls endemic and widespread. The models, however, need to be evaluated because they lack empirical scrutiny and rely on an unverified assumption of widespread resource marginality in dry climates. This evaluation is important because the models and associated expectations have had a profound influence on how we think about the challenges and opportunities of living in dry climates.
Neither I nor the researchers that have employed the models I evaluate assume that demographic and environmental conditions are the only conditions that contribute to vulnerability to dry periods. These conditions and changes in these conditions, however, are often emphasized (individually or in combination) and form an important link in the chain of causal argument regarding climatic influences on human behavior. The individual contribution of a specific condition to dry-period vulnerability has been argued because in the context of assumed resource marginality, changes in either resource demands (represented by demographic conditions) and/or changes in resource supplies (represented by environmental conditions) may have created population-resource imbalances that increased vulnerability to dry periods. The indicators I use to represent differences in these conditions (settlement population levels, watershed population density, settlement locations relative to perennial rivers, and average annual precipitation levels) are also only a subset of characteristics that may influence resource demand and supply.
I identified these models by considering arguments, expectations, and assumptions regarding climatic influences on human behavior prevalent in U.S. Southwestern archaeological studies and in modern studies of vulnerability to natural hazards conducted by researchers in other disciplines. This is the first archaeological effort in the U.S. Southwest or elsewhere, to my knowledge, to explicitly identify these models and systematically examine over an extended period the influence of demographic and environmental conditions on vulnerability to dry periods in order to test the validity of these models. It is important to evaluate these models because they lack empirical scrutiny, are seldom identified explicitly but are rather embedded in other arguments, and rely on an unverified assumption of widespread resource marginality due to low precipitation in the region.
In this chapter, I describe these models and provide examples of their use and discuss how they have affected interpretations of vulnerability to dry periods as well as important events in the prehistory of the U.S. Southwest. The models are summarized in Table 3. I also provide an initial description of how I evaluate each model with greater detail provided in Chapter Five.
See Table 3.1.
Aridity Model of Vulnerability to Dry Periods
Aridity models emphasize the contribution of low precipitation and resource marginality to vulnerability to dry periods. Vulnerability in this model is understood in terms of the biophysical conditions of the environment (resource marginality and aridity), ignoring characteristics of human populations (Liverman 1990b). Biophysical vulnerability "is a function of the frequency and severity (or probability of occurrence) of a given type of hazard" (Brooks 2003:4) such as a dry period. Social, demographic, and other conditions of human populations exposed to dry periods and the contribution of these conditions to their vulnerability to dry periods are either ignored, not emphasized, or assumed to be small in biophysical models of vulnerability to natural hazards.
An emphasis on the contribution of biophysical conditions to vulnerability to climatic hazards is widely applied by engineers and economists in the technical literature on disasters (Fussel 2007) and within global environmental change studies (Liverman 1990b:30). An emphasis on biophysical conditions has its theoretical origins in the risk-hazard vulnerability approach originating in geography. Two examples of this approach are vulnerability to sea level rises and vulnerability to floods. Topographic contours are used almost exclusively to identify the extent of vulnerability to these hazards (e.g., Titus and Richman 2001). Regarding vulnerability to global environmental change, a biophysical approach "implies that in order to understand and delimit vulnerability, we just need to know how and where the physical environment will change. Physical indicators will then provide adequate insight into the populations at risk" (Liverman 1990b:30).
Archaeological acceptance of a biophysical model of vulnerability to dry periods (and other climatic conditions) is pervasive in the U.S. Southwest because of the assumption of resource marginality and widespread vulnerability to relatively low precipitation in the region (as I discuss in the "Resource Marginality” section in Chapter Two). The assumption of resource marginality when employed to explain social and cultural changes in the U.S. Southwest essentially draws the contours of vulnerability to dry periods around the entire region due to relatively low and variable precipitation. As a result, studies of human-environment interactions and climatic influences on human behavior in the U.S. Southwest are usually studies of risk and the contribution of marginality and climatic event severity to this risk (e.g., Doyel and Dean 2006; Graybill et al. 2006; Gumerman 1988; Larson et al. 1996; Minnis 1985; Nials et al. 1989; Rautman 1993; Tainter and Tainter 1996).
Studies that emphasize quantitative estimates of precipitation levels relative to "normal" or average conditions and compare severe conditions to coincident behavioral changes signal the use of an aridity model. For example, Stahle et al. (2007) and Cook et al. (2007) use tree-ring reconstructed summer Palmer Drought Severity Indices (PDSI) to identify decadal droughts more severe and prolonged than any witnessed during the modern instrumental period. Cook et al. (2007; and references contained therein) identify temporal correlations between severe dry periods and a number of important socio-cultural events including the 14th century decline of complex Mississippian chiefdoms and famine, disease, and village abandonment in the Puebloan region of New Mexico in the 16th century. Stahle et al. (2007:136-140) also use the temporal coincidence between identified droughts and historical information on the environmental conditions and activities of Euroamerican and Native American societies throughout the U.S. to validate the PDSI reconstructions. Although Stahle et al. (2007:136) find a number of temporal correlations between severe dry periods and social and environmental events, they also find a number of "disagreements" between these dry periods and known historical events. Stahle et al. (2007:136) attribute these disagreements to climate proxy reconstruction errors or the fact that “tree-ring data integrate moisture conditions during and sometimes preceding the growing season, and may not well represent fall or winter conditions.” I suggest that these disagreements between dry periods and known historical events also strongly suggest variation in human vulnerability to dry periods driven by other factors in addition to weaknesses in the climate reconstruction's ability to accurately identify relevant precipitation conditions and their severity.
Explanations of regional-scale abandonments that attribute them to catastrophic climate events, such as dry periods, treat vulnerability as a regional-scale biophysical condition and assume endemic shortfalls and widespread vulnerability due to low precipitation. This assumption of widespread vulnerability ignores or over-rides any intra-regional social and environmental diversity that influences vulnerabilities. For example, abandonment of the San Juan Basin in the mid-1100s and the Mesa Verde region in the late 1200s has been (Douglass 1929; Judge 1989, and others) and continues to be (Benson et al. 2007) attributed to severe dry periods. Differences in vulnerability across these regions due to varying demographic and environmental conditions and how these differences might have affected dry-period impacts and responses are often not systematically considered. Van West’s (1994) spatially detailed study of soil types and modern crop yields and potential variation in impacts of the late 1200s drought on agricultural productivity is an early and notable exception (see also recent work by Axtell et al. 2002, Kohler et al. 2007, and Schollmeyer 2009). In central Arizona, dry periods combined with catastrophic flooding and associated stream channel changes along the Lower Salt River in the late 1300s have been hypothesized as the cause of the depopulation of settlements irrigating from the Lower Salt River (Graybill et al. 2006, Nials et al. 1989; Gregory 1991). Again, differences in vulnerability based on settlement locations (e.g., near or distant from the floodplain) and settlement-scale demographic conditions that affected the demand for resources are typically not considered.
Dry-period influences on regional abandonments have been argued for many places and times (e.g., Weiss and Bradley 2001). Examples include the Maya of Mesoamerica ca. A.D. 1200 (Gill 2000; Hodell et al. 1995), Cahokia and associated settlements of the American Bottom and Mississippi River Valley ca. A.D. 1100 to 1245 (Benson et al. 2009), the Tiwanaku of Bolivia-Peru ca. A.D. 1100 (Binford et al. 1997), and the Akkadians of Mesopotamia ca. 4200 B.P. (Weiss et al. 1993), to name only a few. Dry periods are seldom considered the only factor responsible for these regional depopulations; however, variability in vulnerability across a region is not addressed in aridity models that emphasize widespread vulnerability to dry periods due to low average precipitation conditions.
To assess the utility of an aridity model for explaining vulnerability to dry periods, I examine the relationship between dry-period severity and residential abandonment throughout central Arizona from 1200 to 1450. I use the relationship between dry-period severity and residential abandonment as an indicator of the extent of vulnerability to dry periods within six watersheds with a long-term history of occupation. By geographically disaggregating the study area into watersheds, I test the assumption of vulnerability to dry periods as a regional-scale, biophysical condition. The aridity model and the assumption of regional-scale resource marginality would be supported by strong and sensitive relationships between dry-period severity and residential abandonment in all watersheds.
Demand Models of Vulnerability to Dry Periods
Demand models emphasize the contribution of demographic conditions to vulnerability to dry periods. The demographic conditions I consider are settlement population levels and watershed population density. These conditions define, in part, resource demands, the rate of consumption of resources, and the extent of labor available to invest in strategies to manage shortfalls. Settlement population levels reflect the cumulative decisions of people to stay or leave a locale--a settlement or watershed area--and to allow others to move into that locale. Exploration of the role of population levels and density is a window into the role of human decisions in creating or ameliorating vulnerabilities to dry periods.
The effect of settlement population levels and areal population density on vulnerability to dry periods is complex and has been argued to both increase and decrease vulnerability (Meyer et al. 1998:241). For example, as population levels increase more resources are consumed and increases in production may not be able to keep pace (Malthus 2001 [1798]) especially during dry-period declines in productivity. Relatively high population density in an area can also increase vulnerability by limiting mobility as a strategy to manage shortfall risks (Binford 1983; Cordell 2000:183; Dean et al. 1994:85; Minnis 1996; Powell 1988). Population density can limit mobility if productive locations are already claimed, occupied, or hostilities restricted movement. Larger populations might also increase the tilling of relatively less productive plots of land to increase production (B. Nelson et al. 1994:61-62) and these plots could quickly become unproductive during dry periods. Thus, "as population density increases, more individuals are forced into areas of greater risk" (Reycraft and Bawden 2000:2). Larger populations can, however, intensify production and invest in infrastructure to increase resource supplies (Boserup 1965) and develop social strategies not as readily available to smaller populations, such as centralized leadership and alliances (McGuire and Saitta 1996; Wilcox et al. 2001a; 2001b), to manage shortfall risks.
Demand models are essentially 'population pressure' models wherein increasing population levels are considered responsible for some stress on humans or the environment (e.g., Smith 1972). Population pressure arguments assert that rising population levels at some point breach a threshold and behavioral responses become necessary. Where resources are assumed to be marginal, this threshold is expected to be easily and frequently breached. For example, Larson et al. (1996:219) assert "With the shift to primary reliance on domesticated crops, population levels and densities increased dramatically between A.D. 900 and 1100 and the Anasazi [northern Southwest] became increasingly vulnerable to climatic variability and extremes." Larson et al. (1996:218) describe this increase in vulnerability in the context of an "extremely marginal environment for prehistoric hunting and gathering and agricultural pursuits." Similarly, Plog et al. (1988:261) state "It is difficult to imagine environmental deterioration of sufficient severity to stimulate a long-distance move in the absence of high populations." High population levels, then, in these examples are a critical factor in understanding vulnerability to dry-period risks of shortfall.
It is important to empirically evaluate this model because expectations of the contribution of demographic conditions to vulnerability to dry periods (and other hazards) are often conflicting (as discussed above) or too ambiguous to be useful or convincing. For example, in a list of possible determinants of vulnerability to global environmental change (which includes increases in dry-period severity associated with global-scale climate change), Liverman (1990b:33) asserts "High population densities, population growth rates, and pressure on limited food, land, and water resources can make regions very vulnerable to global change." Similarly, Reycraft and Bawden (2000:2 citing work by White 1974 and Burton et al. 1978) state that "the greater the population density of a given area, the greater the damage potential of a given extreme event. Also, as population density increases, more individuals are forced into areas of greater risk." And, Cordell (2000:182), referring to upland areas of the northern Southwest, states "Population 'packing' [makes it] impossible to implement an agricultural strategy that depends on relocating fields and dwellings" thereby linking increases in population densities with agricultural vulnerability. These expectations and assertions though sometimes vague are logically appealing and amenable to empirical testing. Minimally, we need to understand at what spatial scales the suggested relationships might apply.
Investigations of the potential influence of increasing population levels on vulnerability to dry periods without considering the potential resource productivity of an area are unrealistic. Coupled with the assumption of region-wide resource marginality, however, these studies place population levels close to a threshold where changes in either population or resources may increase the risk of shortfall. Under this model, people living in settlements or watersheds with relatively high population levels would be more vulnerable to dry periods than those living in places with lower population levels. Thus, for the central Arizona study area, if vulnerability to dry periods was influenced by settlement population levels and watershed population density, the relationship between dry-period severity and residential abandonment will be stronger and more sensitive where population levels and density are highest and weaker where population levels and density are lowest. Results will clarify the influence of these demographic conditions on vulnerability to dry periods in the study area.
Supply Models of Vulnerability to Dry Periods
Supply models emphasize the contribution of the potential productivity of settlement locations to vulnerability to dry periods. The environmental conditions I consider are settlement locations relative to perennial rivers and areas of low to high average annual precipitation levels. Access to resources, as reflected by settlement locations, is one factor that affects the ability of human systems to adapt to and cope with dry periods and climatic conditions. Settlement locations reflect the decisions of people regarding where to live; thus, exploration of the role of settlement locations is a window into the role of human decisions in creating or ameliorating vulnerabilities to the risk of shortfall.
Settlement locations adjacent to perennial riverine resources, which include the associated riparian and aquatic resources, are assumed in this study and model to offer greater potential productivity than settlements located away from perennial riverine resources. The majority of wild plants used as resources by the Hohokam "are most densely and continuously distributed along riparian corridors" (Fish and Nabhan 1991:42). Agricultural potential is also greater in riverine areas where irrigated and floodplain agriculture are possible.
Likewise, settlements in areas receiving relatively more precipitation on average are assumed to have been relatively more productive than those areas that receive less. Differences in precipitation conditions are often used to explain settlement location shifting motivated by changing climatic conditions (e.g. Dean 1988; 1996). Similar assumptions regarding proximity to rivers and precipitation levels have been used to identify "agricultural primeness" in the arid regions of Africa (Miller et al. 2002). Other factors such as the extent of arable land or the quality of soils (Sandor et al. 2007) influence productivity but are beyond the scope of this study.
The influence of potential resource supplies on vulnerability to dry periods wherein vulnerability is considered a function of riverine proximity or precipitation levels is a common sense notion supported by the strong relationship between the distribution of water and settlement locations in dry climates. People living distant from perennial rivers rely on "dry-farming" or "rain-fed" farming and are widely assumed to be among the most vulnerable and sensitive to low precipitation in dry climates (e.g., Liverman and Merideth 2002:207). For example, Harry (2005:299) compares two settlement areas in the Tucson Basin (southern Arizona): one bordering a primary river with wide expanses of arable floodplain, the other along lesser watercourses and in an upland area. Harry (2005:299) "intuitively" expects that the settlements near the primary river were not agriculturally impoverished relative to the settlements along the lesser watercourses and in the upland area. In colonial Mexico, Florescano (1980, as translated by Liverman 1990a:50) argues that "the disastrous effects of drought, as in earlier times, are concentrated in the rainfed agriculture practiced by the poorest ejidatarios and campesinos, lacking credit, irrigation, fertilizers, and improved seeds."
The notion that riverine areas were less marginal than non-riverine areas and people living near perennial rivers were less vulnerable to dry periods than those living distant from perennial rivers lacks empirical verification and there are reasons to question its validity. For example, investments in irrigation infrastructure in riverine areas will increase productivity and resource supplies thereby potentially reducing vulnerability to dry periods. However, these investments in irrigation infrastructure may increase population, reduce mobility, and thereby increase vulnerability to dry periods as well as other social and environmental conditions (Anderies 2006; Ingram 2008; Janssen and Anderies 2007; Nelson et al. 2010). In a study of the impact of irrigation on drought losses in Mexico, Yates (1981) finds little support for the expectation that irrigation has the advantage of reducing climatic hazards compared with crops dependent on erratic rainfall. Rather, he finds "only a slightly smaller deviation from [agricultural] output trend in irrigated than in rainfed areas” (Yates 1981:77 as quoted in Liverman 1990a:58).
To assess the utility of a resource supply model for explaining spatial differences in vulnerability to dry periods in central Arizona from 1200 to 1450, I examine the relationship between dry-period severity and residential abandonment among rooms near and far from perennial rivers and in areas of low, moderate, and high precipitation. If vulnerability to dry periods was strongly related to the productivity characteristics of settlement locations as identified by differences in the availability and access to water, then the relationship between dry-period severity and rooms abandoned near perennial rivers or in areas with high average annual precipitation will be less sensitive and weaker than the relationship between dry-period severity and rooms abandoned distant from perennial rivers or in areas with low average annual precipitation.
Demand and Supply Models of Vulnerability to Dry periods
Demand and supply models emphasize the contribution of population-resource imbalances (at various spatial scales) to vulnerability to dry periods (Cordell and Plog 1979; Dean 1988; Dean 1996; Larson 1996; Larson and Michaelsen 1990). Population-resource imbalances were "probably a fact of life for most prehistoric groups, both because they sometimes grew too fast and because of unforeseeable decreases in the resources available to them" (Cordell and Plog 1979:411). Emphasis on population-resource imbalances are not limited to climate-human behavior studies. Rather, "Interpretations of Southwestern cultural patterns are frequently based on the assumption of high population density, low plant and animal biomass [resource marginality], and resultant population/resource imbalances" (Powell 1988:182).
Attempts to consider both potential resource supplies and demand to infer shortfall risks pose a substantial challenge in archaeological research. It is a challenge because we have inadequate information about the plethora of variables involved in inferring both supply (wild and cultivated food resources) and demand (population levels). For example, identifying the extent of a food supply involves an understanding of the water requirements of prehistoric maize varieties, the proportions of cultivated and wild foods that comprised diets, yields of cultivated and wild foods, and many other factors (Minnis 1985:99-155 provides an extended discussion of these problems).
Despite these challenges, reasonable attempts have been made that provide approximations of periods when population-resource imbalances likely resulted in increased vulnerability to food shortfalls. One such effort is a study developed by Minnis (1985) for the Mimbres region of southwestern New Mexico focused on the A.D. 600 to 1250 period. Minnis documented potential resource supplies over time with estimates of crop success, wild food productivity, and food stress using precipitation and streamflow reconstructions. He compared these estimates to potential resource demands identified by variation in population levels. The resulting effort identified periods when population-resource imbalances and associated vulnerability to dry periods were most likely.
Examining the influence of population-resource imbalances is important because it allows us to see if these imbalances emerge as a consistent influence on vulnerability to dry periods over time. As I discussed in Chapter Two, people have a variety of strategies for managing shortfalls risks and their vulnerability to dry periods. The effectiveness or changes in the effectiveness of these strategies might be a more important factor in vulnerability to dry periods than simple formulations of the potential balance between resource supply and demand. As stated by Dean et al. (1994:86), "Many congruences [between past environmental variability and regional demographic trends] establish the importance of demographic and environmental variables as integral factors in sociocultural stability, variability, and change in this harsh and variable region. On the other hand, many failures of the archaeological record to fulfill the expectations of the models indicate that the effects of population and environmental fluctuations were mediated and transformed by sociocultural factors." Evaluating models of vulnerability to dry periods that emphasize population-resource imbalances, then, help us understand the extent to which these imbalances consistently influenced vulnerability to dry periods.
To assess the utility of this model for explaining differences in vulnerability to dry periods, I combine elements of the demand and supply models (above). I test the expectation that the extent of population-resource imbalances influenced vulnerability to dry periods by comparing the strength of relationships between dry-period severity and residential abandonment where demands were relatively high (settlements located in watersheds with high population density and settlements with high population levels) and supply low (settlements located distant from perennial rivers and in areas of low precipitation) to the relationship where demands were relatively low (settlements located in watersheds with low population density and settlements with low population levels) and supply high (near perennial rivers and in areas of high precipitation). If vulnerability to dry periods was influenced by population-resource imbalances, then the relationship between dry-period severity and residential abandonment will be stronger and more sensitive where demands were high and supplies low (population-resource imbalances most likely) than where demands were low and supplies high (population-resource imbalances least likely).
Conclusion
In this chapter, I describe four conceptual models that emphasize the influence of aridity and demographic and environmental conditions on vulnerability to dry periods. These models reflect logical relationships between the demand and supply of resources and vulnerability to dry periods in areas where resources are considered marginal and the risk of food shortfalls endemic and widespread. The models, however, need to be evaluated because they lack empirical scrutiny and rely on an unverified assumption of widespread resource marginality in dry climates. This evaluation is important because the models and associated expectations have had a profound influence on how we think about the challenges and opportunities of living in dry climates.