Population Estimates and Settlement Patterns

Boundary walls and streets still circumscribe and meander around residential groups at Chunchucmil, giving eloquent testimony to the fact that the city was an integrated and contemporaneous unit without significant disturbance after abandonment (Figure 2). This pattern can be dated to the late Early Classic–early Late Classic periods (a.d. 400–700, or what we call the Middle Classic period; see Varela 1998; Varela and Braswell 2003), based on ceramic and architectural dating of 36 of the hundreds of currently mapped residential groups.

Central Chunchucmil. Note the density of residential groups defined by albarradas (stone fences) and streets, and the central marketplace.

David Vlcek et al. (1978) reported an extraordinary settlement density, but our more recent detailed mapping has shown his estimates to be conservative. These data will be reported on in detail elsewhere, but a preliminary summary here will be useful (Hutson et al. 2006). By the end of the 2004 field season, we had mapped a total of 7.3 km² of the central area of the site, including 1 km² defined as the site center or urban core. For this paper, we have calculated population estimates based on house-mound counts within a 1.84 km² sample area, which we feel is representative of the central 11 km². Aerial and ground reconnaissance indicate a small drop in structural density in a 10 km² urban zone beyond the extremely dense, central 11 km²; because settlement within this 10 km² band is still densely urban, we feel it is appropriate to make an educated guess at its resident population. We have done so by conservatively and preliminarily cutting in half the number of structures and residents in the central 11 km².

We discounted obviously nonresidential structures in our structure counts (e.g., temple pyramids, small ritual platforms in the center of patios, ballcourt, and so on) and ancillary buildings in house lots (e.g., storage buildings, kitchens). Moreover, we have used correction factors (Table 1) in an attempt to eliminate structures known, or suspected, to be later, such as some platform groups that appear to be Late Terminal Classic in date or other non-contemporaneous structures with the site's main occupation. After the application of these conservative correction factors, residential structural density in our sample area amounted to 530 to 595 structures/km² (Table 1). Using an average family size of five persons (a commonly used factor in the Maya area; see, e.g., Bequelin and Michelet 1994; Rice and Culbert 1990), these numbers result in a population density of between 2,650 and 2,975 persons/km² within the inner 11 km² and 1,325 and 1,487 persons/km² in the outer 10 km². Thus, the total estimated population, is between 29,150 and 32,725 within the central 11 km² (2,650 × 11 = 29,150 and 2,975 × 11 = 32,725) and between 13,250 and 14,870 in the outer 10 km² sample area (1,325 × 10 = 13,250 and 1,487 × 10 = 14,870), for a total within the 21 km² provisional area of the city of between 42,400 and 47,600 people. It should be noted that the site is a lot larger than 21 km², but we have no way of quantifying it now. However, using analyses of aerial photographs, Landsat 7 satellite images, and limited ground surveys, we are beginning to recognize residential peripheries that might encompass yet another 29–49 km² (Mazeau and Hixson 2005).

Structure Density At Chunchucmil

Land Required for Agricultural Self-Sufficiency

Regardless of the final population estimates, our preliminary figures of 42,400–47,600 people are sufficient to suggest that regional agricultural self-sufficiency using contemporary maize-production and maize-consumption figures is problematic. As a consequence of the poor precipitation and soil conditions (see later), average milpa production today in the Chunchucmil region is very low—about .25 metric tons/ha without the use of fertilizer and up to 1 metric ton/ha in an ideal year (see Beach 1998a). These yield estimates are in basic agreement with those from Komchen (without fertilizers), which also has deficient soils and rainfall (Shuman 1974). Here yields are said to be about .25 metric tons in a good year, and in mediocre years yields can be as low as .1m ton/ha (Shuman 1974:158). To put these yield estimates in perspective, we offer the following comparisons: average world yield equals 4.6 metric tons/ha (Pinstrup-Andersen 1994); yields near Lamanai in northern Belize with deeper soils and plentiful rainfall equals 1.7–.8 metric tons, depending on how long milpas were worked (Arnason et al. 1982); two-year average yields in Xocen, Quintana Roo, also with deeper soils and plentiful rain but with moderate use of fertilizer is less than 1 metric ton/ha (Terán and Rasmussen 1994); and yields in the Piste region near the center of the northern peninsula, with deeper soils and more precipitation, was about.75 metric ton/ha (Steggerda 1941).

Under subsistence conditions, a Maya family of five consumes a little more than a metric ton of maize per year at Piste and Komchen (Shuman 1974; Steggerda 1941). Based on long fallow swidden agriculture yields of .1–1 ton/ha, this family's annual food requirements, therefore, are between 1 ha and 10 ha actually under production, which leaves precious little margin of safety. With an average solar size of 4,233 m², or .42 ha, available to residents in the 21 km² core and residential districts, only a tiny fraction of a household's maize requirements could be produced, and this does not include the spaces occupied by buildings and nonagricultural activity areas estimated by Thomas Killion (1992) at 60%. With a population of 8,480–9,520 families (42,400–47,600 persons divided by an average family size of five persons), urban Chunchucmil would have needed a total of 8,480–95,200 ha, or 84.8–195.2 km² of maize under production in outfields in any given year. However, fields can only be used for about two years without significant yield declines due to soil-nutrient depletion and weed competition (Beach 1998b). They are left fallow for 10–12 years, though fallow times have recently shortened with the advent of chemical inputs. Using chemical inputs and assuming contemporary indigenous 10–12 year fallow practices, then the 8,480–9,520 families within the 21 km² sample area would have needed a total of 424–1,171 km² of cultivable land available to them (Table 2).

Preliminary High and Low Estimates of Land Required to Sustain Chunchucmil's Urban Population Based on Long Fallow

Whether that amount of land was available in the region is problematic. First, Chunchucmil is bounded by an agriculturally unproductive savanna that fingers toward it to the west; Oxkintok's sustaining area, which begins fewer than 16 km to the east-southeast (Misión Arqueológica de España en México 1987–1990); and Siho's sustaining area 16 km to the south (Dunning and Andrews 1994). The savanna to the west is inundated up to a meter deep in some places during the growing season, and this prohibits the cultivation of staple food crops. A few narrow fossil beach ridges (ca. 1–1.5 m high) known locally as tzekeles present few and relatively tiny viable planting and habitation surfaces here. Moreover, there are no vestiges of wetland agriculture in the freshwater swamps immediately to the west of the savanna. We have therefore excluded all wetlands from our estimate of the agriculturally productive land in what we had initially defined as the Chunchucmil Economic Region or sustaining area (see Dahlin and Ardren 2002). We currently estimate the maximum extent of the surrounding arable land at about 1,600 km², but a considerable overlap with Oxkintok's and Siho's sustaining areas is contained within that 1,600 km² area using a Thiessen polygon as a guide to drawing site boundaries. Therefore, we do not know the precise size of Chunchucmil's sustaining area, but it must be significantly less than 1,600 km². It is clear, even at this juncture, that only the high-end estimate using the shortest fallow cycle and the highest yields can accommodate even our lowest population estimate.

Second, our Classic-period population estimates leave out the 29–45 km² of the residential periphery. It seems plausible to us that not even the 10 year fallow cycle that is common today could have sustained the entire population within the combined urban core, residential zone, and periphery, and for reasons stated later, the fallow cycle could not be shortened much beyond 10 years.

Third, with virtually no land pressures exerted by extremely low population densities today relative to those projected for Chunchucmil's population climax, contemporary farmers have more options and can be selective about where they place their milpas. For example, they can avoid planting on large portions of the land covered by limestone pavement and the very thinnest soils, and they can plant in the deep, fertile soils developed over the last 1,300 or so years on abandoned Classic-period building slopes. Therefore, productivity of milpas strategically placed to maximize fertility and soil depths today greatly exaggerates Classic-period productivity.

It should be clear from the foregoing that the nature of the science makes any estimates of population and production broad. However, our estimates are based on the most conservative demographic and production figures. While these estimates are far from conclusive, they are sufficient at this stage to sustain the hypothesis of agricultural insufficiency based on contemporary milpa practices. They also compel us to examine factors that would influence strategies of agricultural intensification.

Precipitation

Successful cultivation of maize requires a minimum of about 500 mm of rainfall during its growing period (Wellhausen 1957). A total of 640–900 mm is available in the study area in an average growing season, but a high percentage of that is lost in karstic seepage through the permeable limestone bedrock and in evapotranspiration, resulting in a mean annual water-budget deficit of 600–700 mm (Instituto Nacional de Estadística, Geografía e Informática 1983; Luzzadder-Beach 2000). Also, a high percentage of the rainfall occurs between May and September, which limits production to one crop per year.

The convectional precipitation regime is advected by southeasterly trade winds bringing moisture in from the Caribbean Sea westward across the peninsula. Satellite imagery thus reveals that the western and northern portions of the peninsula are in the rain shadow of the eastern peninsula, which depletes rain clouds increasingly as they move westward (Williams 1976). Thus, the Chunchucmil Economic Region receives less benefit from the southeasterlies than regions to the east and south.

Just as important as the amount of rainfall are its regularity and predictability, which affect clearing and burning, seed germination and maturation, and harvesting of crops. It is well known that precipitation on the northern peninsula is unpredictable in its occurrence because of highly localized convectional thunderstorms, droughts, and hurricanes. Some years have three to four times more rain than other years (Lundell 1934), and modern figures show variations by as much as 30% from the long-term average (Instituto Nacional de Estadística, Geografía e Informática 1981). Some of this variability is due to hurricanes, which recur on average every eight to nine years. However, the ethnohistoric and historic literature is full of references to drought (see Gill 2000 for the most complete discussion of droughts). Particularly troublesome in some years is a long, often devastating canicular drought in the middle of the growing season that exaggerates year-to-year variability and growing conditions. Paleoecological indications in lake-core sediments 30 km north of Chunchucmil reflect essentially the same climatic conditions throughout the Classic period (Brenner et al. 2000; Leyden et al. 1996). Particularly sensitive lake-core data from the eastern side of the peninsula show frequent and enormous fluctuations in precipitation (Curtis et al. 1996). Thus, rainfall amounts and variability today—and in the past—are difficult for maize. Moreover, we might expect the extreme aridity of the region might select against some of the more hydric and mesic cultivars used by the more southerly Maya and select for cultivars that can withstand prolonged drought. As we will see later, only a few of the latter cultivars have been seriously entertained as staple foods.

Maize yields could still have been bountiful with enough irrigation, but two of the most serious factors limiting settlement on the northern plains are water accessibility and quality. There are no rivers or lakes in the region, and rainwater percolates rapidly down through the top 2–4 m of porous, fractured bedrock to a mounded water table. Ironically, too much water hampers crops in places where clay plugs limestone pores and water ponds up for several weeks after large storms, which are frequent during the growing season.

Geomorphic Features

The region is a remarkably flat karstic plain wedged between the Sierrita de Ticul on the east and broad savanna and coastal swamp/estuary zones to the west (Figure 1). A cap rock, about 1 m thick, covers a fractured and extremely porous Quaternary and Pliocene limestone, which is laced with veins and pockets of sascab (a soft, crumbly limestone) (Dunning et al. 1998; Perry et al. 2003). Bedrock outcrops, or uitzes, pockmark the landscape to the north, east, and south of the site and increase in micro-elevation from west (ca. 1 m) to east (ca. 3 m). These uitzes are highly fractured and contain relatively deep pockets of richer, more water-conserving soils (see later).

Depressions, or dolines, are fairly common, although cenotes open to the water table are exceedingly rare in most of the study area. Instead, the landscape is dotted with shallow (ca. 1–3 m deep) rejolladas, or dry solution features. The rejolladas, as well as sascaberas, contain rich organic soils but are generally too small to provide much in the way of bulk staples.

Natural drains measuring just a few centimeters in diameter are common. While they are visible only on bare limestone pavements, they are probably much more numerous. Although large expanses of ground support puddles and even ponds for some time after rains, unseen drains conduct an unknown, but probably voluminous, quantity of surface water down to the water table. Larger (ca. 1–1.6 m in diameter and perhaps as deep as 90 cm) swallow holes collect some rainwater for weeks after heavy rains. They look deceptively like wells or cisterns. They, like the drains, would have provided deep, moist soils for trees such as ramon or brosimum alicastrum.

Hydrology

Water accessibility and quality are two serious factors limiting settlement on the northern plains. There are no rivers or lakes in the region, and natural and easily accessible water sources are rare on the cultivable portion of the region, although they are quite common in the uncultivable savanna (see later). Two reservoirs about 250 m apart are located in the epicenter, but they are clearly too small and their waters too ephemeral, even at the height of the wet season today, to have been secure sources of potable water, much less sources for pot irrigation. Chultunes, or cisterns, that are so common elsewhere in the northern Maya Lowlands (Matheny 1971; McAnany 1995) are unknown. We have found a total of 20 modern and ancient Maya wells, using informants within about a 5 km radius of the site center, and four wells beyond that. We excavated three wells in or near the site center (Dahlin and Jones 1998). In addition to the two reservoirs in the epicenter, four wells are known in the civic or communal spaces; the rest occur within residential units.

With a land-surface gradient from sea level at the coast to Chunchucmil about 27 km to the east less than 20 cm/km (Perry et al. 1989, 1995, 2003), the water table is quite high—some 2–3 m—and not much deeper elsewhere in the immediate vicinity (Luzzadder-Beach 2000). Wells are easy to dig in the soft limestone under the cap rock (Ringle and Andrews 1988); in fact, they are much easier to dig than chultunes (see Matheny 1971). The depth to the water table might also have been lessened by locating wells in rejollada floors, large natural drains or sascaberas (gravel or sascab quarries dug just under the cap rock into the walls of rejolladas). However, we have not seen evidence of wells in rejolladas floors and natural drains, of which three of the latter were excavated. Sascaberas sometimes retain a few centimeters of rainwater for weeks after large storms, but the availability of these waters is too unreliable during normal years, let alone during the common drought years. Moreover, only about 25% of households had sascaberas. Nevertheless, sascaberas were obviously of great importance to ancient Maya families insofar as all of them so far observed are walled off from public access by dry-laid stone walls or albarradas (Magnoni 1995). In almost all cases, the undermined cap rock has collapsed, which presented a formidable risk factor in antiquity and has deterred us from further testing. One villager in Chunchucmil Pueblo, however, found a well at the bottom of collapse debris in an ancient Maya sascabera. Today he uses it to pump water to a fine huerta (orchard) as well as for his family's potable water.

There probably were more wells than are currently accessible, since some wells may be obscured by collapse debris and others deliberately covered to prevent children or livestock from falling in them. Nevertheless, after having mapped 7.3 km² in great detail and paying considerable attention to water accessibility, we remain perplexed by the paucity of water sources, given the high precipitation deficits in the region and the high population estimates.

One possible solution to the dilemma may rest on developing a sea-level curve on the nearby shores of the Gulf of Mexico. Rainwater percolates down to a mound of freshwater that floats on a deep bed of salt water (Back and Hanshaw 1976; Luzzadder-Beach 2000; Perry et al. 2003). This freshwater mantle is well connected to the Gulf of Mexico via caverns inasmuch as fluctuating water levels in wells have been observed throughout the day and night. This articulation of the aquifer with the Gulf of Mexico means, in turn, that depth to the water table at any point on the landscape will fluctuate with sea-level changes.

The question of sea-level changes is potentially an important one. Karen Walker and her colleagues (Walker 2000a, 2000b; Walker et al. 1994) have found evidence from Florida of rising levels from about 60 cm below pmsl (present mean sea level) circa 1050 and 50 b.c. By about a.d. 250–500, sea level began to crest at perhaps as much as 137 cm above pmsl. This was followed by an interval of falling sea levels to about 60 cm below pmsl between circa a.d. 500 and 750. W.F. Tanner (1992) and F.W. Stapor and colleagues (1991) both report another rise in sea level and transgressions in the Gulf of Mexico in the ninth–eleventh centuries. We are compiling evidence that the Florida sea-level changes approximate in time, direction, and magnitude those recorded on the coasts of the Yucatan Peninsula (also see Dahlin et al. 1998; McKillop 1995). We do not want to argue the applicability of these proxy data here other than to say that they probably result from the same eustatic processes and that the depth of ancient wells in near-coastal areas such as Chunchucmil should mark the prevailing sea stand at the time each was dug.

Water in some of Chunchucmil's ancient wells is currently up to about 140 cm deep; other wells have only about 40 cm of water; and still others have no water in them at all. In all cases, however, there is invariably a small ledge about 40 cm above the floor. This ledge was very convenient to sit on during our well excavations in 1998 and probably was used in the same fashion by the original well digger. No doubt the ledge was kept to help tip vessels while they were being filled; about 40–50 cm of water below the ledge is more than ample for submerging and filling water vessels, and it would be almost impossible to dig much farther without drowning. The elevation of the top of the ledge should therefore reflect the height of the water table. Thus, we believe that those wells with more than 100–110 cm of water above the ledge were dug when the water table was up to 1m below pmsl, while those with a water line very near the ledge were dug when sea level approximated pmsl, and those wells with no water were dug when the water table was higher than pmsl.

Suffice it so say here that high sea stands of this magnitude would force large inland transgressions of the swamp/estuary zones and perhaps the savanna on this exceedingly low and flat coastal plain and this in turn would mean a considerable encroachment on arable land. Conversely, more cultivable land might have been available during periods of low sea stands as these wetlands retreated. The magnitudes of expansions and contractions of cultivable land might well have been significant at particular points in Chunchucmil's land-use history. This possibility obviously deserves more research. Also deserving of more hydrological research is the fact that access points to groundwater currently appear to be too thinly distributed among Chunchucmil's households and outfields to have provided much irrigation, despite the region's marginal precipitation.

Soils

Between 25% and 50% of the ground surface lacks any soil at all, as noted over the years by Bishop Landa (Tozzer 1941:181) to modern commentators (Beach 1998a; Isphording and Wilson 1973; Perry et al. 2003). Where soils do exist in this environment, they are often skeletal, a fact that restricts nutrient uptake and soil-moisture capacity. These soils are often underlain by dense petrocalcic layers, which provoke localized soil flooding and hinder the development of young crops during wet periods. At least 30% of the rest of the landscape is covered by thin, reddish brown, and dense clay soils locally called kancabs (Paleustalfs, Haplustolls, and Haplustalfs in U.S. Department of Agriculture [USDA] terms). Thus, 55% to 80% of the present-day landscape has thin to no soil.

The remainder of the landscape is covered by thicker, more fertile, black, moisture-conserving clay loam, or boxluums (Calciustolls and Haplustolls in USDA terms), that develop on ancient Maya architecture, in fractured cobble and gravel limestone, on uitzes (among the most preferred sites for cultivation for farmers today), and in rejolladas. The young boxluums on mounds and plazas would not have been available to the ancient Maya. Although more boxluum soils are distributed around the site than in the rest of the karst plain, we are uncertain of the proportion that naturally occurs in bedrock fractures and rejolladas, as opposed to that formed over the past 1,000 years on abandoned Maya architecture. But with 55% to 80% of the landscape covered either by limestone pavement or thin kancab soils, the boxluum pockets on broken and pockmarked terrain might have been a pull factor for settlement and a permanent economic advantage to the descendants of the original pioneers.

There is no reason to think these conditions did not prevail in antiquity because the landscape could not have eroded away, as was suggested about this region in the past (cf. Conservation Foundation 1954; Robles Ramos 1950). Unlike the central Maya Lowlands, where steep slopes led to significant soil erosion in the Preclassic through Classic periods, and the Maya developed conservation mostly in the Late Classic (Beach, Luzzadder-Beach, Dunning, and Scarborough 2003; Beach, Dunning, Luzzadder-Beach, Cook, and Lohse 2006; Beach, Luzzadder-Beach, Dunning, Hageman, and Lohse 2002), the Chunchucmil plains are too flat (far less than 1° slopes on average) for much soil erosion by runoff, and the high clay content and vegetative cover preclude much wind erosion. These fundamental factors that make erosion unlikely here are attested by the very small amounts of eroded sediments in local depressions and in the Cenote San Jose Chulchaca core 30 km to the north (Brenner et al. 2000; Leyden et al. 1996).

It is natural weathering and nutrient removal by crops that are the major soil-fertility losses in this landscape. Organic matter, base saturation, cation-exchange capacity, calcium, magnesium, sodium, iron, and nitrogen are generally at healthy levels, but the essential nutrients of phosphorous, potassium, and zinc are variable but generally low (Beach 1998a). Removal of harvested crops from the field substantially diminishes soil nutrients, and soil manipulation can lead to the decomposition of organic matter, a storehouse of essential nutrients. Milpa production in the Maya Lowlands and elsewhere in the tropics is always challenged by weed (and pest) encroachment, especially after the first year of cropping. Indeed, most recent evidence indicates that weed accumulation and the mounting labor required for weeding are the primary causes for abandonment (Beach 1998b; Kellman and Tackaberry 1997:217).

In the lowlands of Belize, John T. Arnason and colleagues have provided a number of baseline studies of plant, soil, and nutrient interactions in shifting cultivation (Arnason et al. 1982; Brubacher et al. 1989; Lambert et al. 1991). They conclude that yield declines were due to weed competition for light and water because by the third year of cropping, weeds had more of all macronutrients than corn. Milpa weeds increased annually, holding more and more nutrients in each year of cropping (Arnason et al. 1982; Brubacher et al. 1989). Fallow leads to the succession of woody plants and trees over herbaceous weeds and diminishes the number of weed seeds over time (weed seeds can stay viable for more than 45 years). Thus, the most important role of fallow is controlling weeds. Lambert and Arnason (1989) found that a two-year cropping cycle with slash and burn was the most efficient use of nutrients and weed competition. The differences between the sites in these studies support the contention by Pedro Sanchez (1976) that milpa abandonment occurs on poor soils because of reduced nutrients but that it occurs on rich soils due to weed competition for light and water (also see Dahlin 1987).

Based on milpa agriculture in Yucatan, sustainable cropping requires either a minimum of 10–12 years of fallow to restore adequate levels of phosphorous, the minimum fallow period practiced by contemporary traditional farmers today (Shuman 1974)—or mulching. Nitrogen is sufficient at most study sites in the region except in deeply buried soils that are removed from the organic cycling and increases by nitrogen fixation in the common leguminous vegetation of the region, to say nothing of the restorative effects of inter-cropping with nitrogen-fixing legumes. Thus, phosphorus values limit the fallow cycle to not less than 10–12 years and impose the strictest limit on agricultural intensification.

Finally, we note that fluctuations in the water table due to changing sea levels may have influenced the amount of arable land that was available, as they have a tremendous potential for expanding and contracting the savanna zone on this exceedingly low and flat landscape. These expansions and contractions in turn make more agricultural land available during periods of low water, such a during the Late Classic period, and less land for cultivation during high0water intervals, such as, perhaps, during the Early Classic period.

Maximizing Ground Space

Scott Hutson and colleagues (2006) selected a sample of 75 residential groups that were completely bounded by albarradas to give an idea of the range and median of solar (yard) size and therefore the amount of land potentially available for outdoor activity areas, including kitchen gardens, urban orchards, and infields. After subtracting the surface area of architecture, solar sizes were seen to range from .075 ha to .75 ha, with a median of .253 ha. Solar size does not correlate with distance to site center, but it does correlate with volume of stone architecture and floor area, both measures of group size. Thus, the more populated residential groups had more potential planting surface, but there were, of course, more mouths to feed. Settlement density and small solar sizes precluded sufficient maize production in Chunchucmil's urban area by a wide margin. However, solares might have allowed some intensive gardening or seedbed cultivation, particularly if all available ground space had been maximized and protected. That they were protected is demonstrated by the sharp definition of solares, gardens, orchards, and infields by albarradas and callejuelas, or streets.

A well-known space-saving intensification technique is polycultural gardening, which attempts to maximize precious land surfaces by distributing crops vertically rather than horizontally (Atran 1993; Barrera 1980; Barrera et al. 1977; Folan et al. 1983a; Killion 1992; Nations and Nigh 1980; Stuart 1993). Such systems include some combination of root crops, such as jicama (Pachyrhizus erosus [L] Urb.), yucca (Manioc esculenta), cocoyam (Xanthosoma sp.), and sweet potato (Ipomoea batatas [L] Lam.) in the subsoil; ground-hugging herbs, vegetables and shrubs in lower stories; shade-tolerant tree species in middle stories (e.g., balche, lonchocarpus); and larger trees in upper stories (e.g., ramon) and dominants (e.g., Folan et al. 1983; Wiseman 1978). These systems combine the strengths and weaknesses of different crops on the same plot to fix nitrogen, conserve soil moisture, distribute sunlight, and minimize pests. Polycultural gardening has received a good deal of agroecological research, which has brought many intensification techniques of this kind to light that were fully within the capacity of the ancient Maya to use (e.g., Gómez-Pompa 1987; Hernandez 1980; Reyes and Sanchez 1990; Terán and Rasmussen 1994; Varguez Pasos 1981; Vargas et al. 1985; Zizumbo Villareal and Colunga García-Marín 1993; Zizumbo Villareal et al. 1992).

Staple polycultural crops, however, tend to be nuts, grains, or roots. Polycultural gardening is best practiced in regions with greater rainfall and deeper soils than are generally found at Chunchucmil, and it is not well developed in the region today (however, see Ortega et al. 1993). Moreover, polycultural gardening is difficult to demonstrate archaeologically because its efficacy is due to judicious use of biotic architecture, which does not preserve well in the archaeological record. Recent pollen evidence at Cobá (Leyden et al. 1998), nevertheless, demonstrates the process of removing maize fields from Cobá's site center during the site's demographic peak in the Late Classic period, a process that can be interpreted as an attempt to maximize the efficiency of urban land use for habitation and associated gardens and orchards.

Ambiguous evidence for arboriculture in Chunchucmil's residential district lies in occasional piles of rough stone, about 3 m in diameter. These stone circles, or “chich mounds,” may be the remnants of “tree foundations” intended to conserve moisture; they are at least superficially similar to walls erected around large trees for that purpose in Yucatan today (Kepecs and Boucher 1996). Rejolladas and sascaberas provide other opportunities for arboriculture. Arturo Gómez-Pompa and colleagues (1990) and Susan Kepecs and Sylviane Boucher (1996) make a good case for growing cacao (Theobroma cacao) in the much larger and more ubiquitous rejolladas, cenotes, and sascaberas along the northern coast, and William Folan (1978:84) reports that the contemporary Maya around Coba intensively garden these features today. In fact, within Chunchucmil, both rejolladas and sascaberas are confined to private yards and are protected from public access by albarradas. Finally, large natural drain holes in solares might have served as “tree planters.” We excavated three such units that had a diameter of about 1 m. Two of the drains contained very little cultural material, a finding that we feel is consistent with their use as tree planters inasmuch as a tree trunk would have blocked the orifice from being used as a midden. The third drain contained abundant cultural material, but it was located at the back of a house that stood on top of a meter-high platform; it is possible that cultural material below such an artifact-rich environment (perhaps a midden) tumbled into the drain sometime after a tree ceased to occupy it. Thus, these features—chich mounds, rejolladas, sascaberas, and natural drains—no doubt provided important economic products for those households fortunate enough to have them.

The most likely candidate for a staple arboricultural crop and alternative to maize is, of course, ramon, even though Chunchucmil lies beyond Dennis Puleston's (1968) zone where ramon thrives due to the aridity and thin to nonexistent soils. This broad-crown canopy tree produces a highly nutritious and storable nut that we know was eaten at least during times of famine (e.g., Alcorn 1984; Atran 1993; Lentz 1993; Puleston 1968, 1982; Tozzer 1941). Fossil pollen record its presence almost throughout the Holocene here (Leyden et al. 1996), and today ramon trees grow in reasonably well-developed regional second growth (mature stands of forest being exceedingly rare), but they appear to be relatively stunted. Tree crops such as ramon can thrive in shallow soils, when they can tap into deeper water sources. They may not achieve the height, maturity, and productivity as in the well-watered south, or along the Paseo Montejo in Mérida, where these large, memorable trees are heavily irrigated. But this requirement does not eliminate ramon from the small list of potentially important food supplements because mature economic trees are almost certainly capable of sending down taproots through the fractured limestone to offset water shortage, or they could be planted in the natural drains that dot the landscape.

Puleston (1982:361) cites one ramon study stating that in better-watered areas of Yucatan and Campeche, a single large tree can produce up to 60 kg of edible seed annually, and in these same areas where ramon leaves provide valuable forage materials for cattle and horses, a spacing of 100–125 trees/ha can be maintained. This spacing would allow some under-story cropping and produce between 5,700 and 10,100 kg of edible seed/ha, which is far in excess of the annual basic subsistence needs of a family of six. With a median of.253 ha, however, most solares at Chunchucmil are too small to contain more than 20–25 broadly crowned trees without shading out many highly valued herbaceous comestibles and nursery crops that require a great deal of sunlight. Nor does it allow for the stunting of all subtropical and tropical tree species in this dry zone of northwestern Yucatan. Nevertheless, under these most optimistic conditions, ramon groves of this size might produce about 1,200–1,500kg/year. Consumption figures are not available for ramon, but if we assume for the sake of argument that ramon and maize are of approximately the same nutritional value (see Puleston 1968, 1982) and that a family of five can subsist on a little more than a metric ton of maize per annum, then it is possible for that same family to subsist totally on a diet of home-grown ramon. Chich mounds, rejolladas, sascaberas, and natural drains no doubt did provide important economic products for those households fortunate enough to have them. Given the prevailing thin soils and generally deficient rainfall over most of the urban landscape, the small size of most solares, and the fact that chich mounds, rejolladas, sascaberas, and natural drains are few in number, and generally small in size, it seems unlikely that ramon could have provided much in the way of bulk food staples.

There are other reasons to speculate that ramon served only as a dietary supplement rather than as a broadly relied on staple at Chunchucmil. First, its botanical remains have never been found at any site other than Cobá (Lentz 1999:13). Second, ethnohistorical analogy does not favor heavy reliance on kitchen gardens for basic staple production. Javier Caballero (1992) writes that only about 11% of the average domestic calorie intake is produced in kitchen gardens. Despite the fact that the historic and modern Maya value their orchards highly, polycultural gardens are normally limited to market products (Herrera 1994:26) and nutritional and dietary supplements, medicinal herbs, and plants used as ornamentals and ritual objects rather than bulk staples such as ramon (Caballero 1992). Third, ramon, as is well known, is, and historically was, held in very low regard as a food stuff (Puleston 1968). Thus, while we can probably eliminate ramon as a food staple for most of Chunchucmil's population, we cannot eliminate it as a food supplement.

Water Management

Pot irrigation remains the most promising water-management practice, given the high water table and the ease of well digging through the soft limestone bedrock. As we suggested earlier, excavation of wells would have been made even easier on sascabera floors, but sascaberas are found in only about a quarter of the residential units we have mapped thus far. Rellojadas also would have lowered the ground surface to make the water table more accessible in wells, and this would have increased the opportunities to pot irrigate urban gardens. Given the difficulty in finding wells on the ground surface and on rejollada and sascabera floors, it is even possible that well water, like sascab (Dahlin and Ardren 2002), was commodified and sold or bartered to neighbors to pot irrigate their gardens. Be that as it may, it seems unlikely that pot irrigation was widely practiced in milpas outside the city, where many wells and huge amounts of labor to dig, maintain, and use them would be needed. Local informants have not reported the existence of wells outside the city on karstic plains to the north, east, and south. However, wells—some modern—abound in the savanna to the west, especially next to tzekeles, where the water table is just centimeters below the ground surface and crops on these micro-topographic highs can be pot irrigated. Alternatively, apiaries are often associated with these water sources; indeed, the abundance of apiaries here is unparalleled.

Water-quality studies revealed that potability for domestic use and consumption is limited by the concentrations of nitrate (N), sulfate (SO₄), and chloride (Cl). Mean Cl levels of 188 mg/l (milligrams/liter) place Chunchucmil's ground-water quality in the “increasing problems” category of toxicity from root absorption (Luzzadder-Beach 2000; Luzzadder-Beach et al. 1986). This means that tree or woody crops are more vulnerable to root absorption toxicity than are annual crops, but high levels of salinity, and total dissolved solids (TDS) limit field crops, as well. TDS and salinity bear close linear relationships to the field measure Electrical Conductivity (EC, measured here as microsiemens or μ). Chunchucmil's ground water EC averaged 1,174 μS, but individual wells ranged from 310 to 3,100 μS (Luzzadder-Beach 2000), which can translate to a 0% to 35% reduction in maize yields due to root absorption and salinity, depending on the individual water source's EC (Luzzadder-Beach et al. 1986, California Department of Water Resources 1978) .

Analysis of ground-water chemistry produced two rather surprising results. First, water quality in terms of N is relatively good in the region, despite numerous point sources for pollution (e.g., corrals and human settlement), probably due to rapid flushing of the aquifer (Luzzadder-Beach 2000). But second, high Cl values may be restrictive with respect to agriculture. Cl values above 142 mg/l indicate increasing problems with root absorption by tree crops, and values above 106 mg/l mean increasing problems with foliar absorption in annual field crops (California Department of Water Resources 1978; Luzzadder-Beach et al. 1986). Fourteen wells showed Cl concentrations exceeding these values. It is not yet known how restrictive these high Cl values are with respect to the kinds of crops the ancient Maya were probably cultivating, but they could be consequential when added to the limitations imposed by soil chemistry and climate.

In sum, while serious rainfall deficits would seem at first glance to present serious limitations to agricultural productivity, pot irrigation from nearby wells could be used in dooryard cropping systems, and tree species with tap roots capable of penetrating the water table might also thrive. But we do not know how ground water's high salinity may negatively affect productivity of annuals and tree crops.

Soil-Fertility Management

Soil organic matter is subject to rapid tropical decomposition and rapid cycling. Burning and harvesting crop plants remove soil nutrients that must then be renewed. The soil nutrients that normally pose the most important limitations to agricultural productivity are phosphorus and nitrogen. Unfortunately for us, there are no direct means of measuring Classic Maya additions of soil organic matter and nitrogen (but see later). Nitrogen replacement was, and is, not a serious problem with fallowing and crop rotation because the low deciduous forest that covers the region has a high percentage of endemic, nitrogen-fixing legumes. In fact, second-growth seres have up to 90% of such species (Flores and Espejel Carvajal 1994:70).

Improving nitrogen can also be done by green manuring or clipping a legume cover crop in milpas and gardens and adding it as mulch. This is essentially the same idea as the underreported but widespread method of mulching called “tapado.” Tapado is the practice of interplanting a green cover crop, usually a legume that will fix nitrogen, add some phosphorous, protect the soil from erosion, ameliorate soil structure and water movement, enhance fertility, and suppress weeds and perhaps pests (Reyes and Sanchez 1990; Thurston et al. 1994). The cover crop is periodically clipped and composted in situ. Agronomists at CINVESTAV (Centro de Investigaciones y de Estudios Avanzados) are currently experimenting with tapado, but the efficacy of this strategy is not clear thus far.

Phosphorus replacement is the single most limiting factor. We can estimate ancient additions of phosphorus by means of phosphate fractionation (e.g., Beach 1998a; Eidt 1985; Lillios 1992). A relative high Fraction 2 to Fraction 1 can indicate phosphate enrichment in the past. Fraction 1 is the most soluble fraction, a nonoccluded fraction bound with iron and aluminum. Fraction 2 measures phosphorus occluded within iron oxides, and hydrous oxides resists weathering, and persists for very long periods of time. At Residential Chunchucmil, likely ancient soils have low Fraction 1 compared with Fraction 2 in an infield, a soil buried under a sacbe, a small sink, and a sascabera. High fractionation ratios elsewhere (Table 3) may indicate that Chunchucmil's soils experienced some kind of ancient phosphorous enrichment. Likewise, the five lowest ratios all occur on what are likely sediments deposited since the Late Classic period and not fertilized by ancient Maya farmers.

Phosphate Fractionation of Chunchucmil Soils (partly based on Beach 1998a)

These findings are very different from the P-fractionations (N = 10) from Sayil (Dunning 1992), where the total phosphorous mean is 538. 9 mg kg-1 (milligrams per kilogram), and the ratio of Fraction 2 to Fraction 1 is .082. The same figures for Chunchucmil are a total phosphorous mean of 340.1mg kg-1, and the mean ratio of Fraction 2 to Fraction 1 is 2.8. Compared with those at Sayil, even the modern used soils with the lowest phosphorous values at Chunchucmil still have high phosphorous-fractionation ratios (.94–1.39), which may be remnants of less intensive historical fertilization or the byproduct of mixing with a highly fertilized ancient landscape.

There are a variety of ways in which the Maya could have improved phosphorus in the soil: mulching with household wastes, composting, and importing soils and organic matter from wetlands and/or burnt sascab. Collection of household wastes might have provided a bit more nitrogen to dooryard gardens, but again, nitrogen is not a serious problem. Night soil and other household wastes were no doubt collected, processed, and spread on gardens, although the quantity would not have made a dramatic difference. Regardless of how effective waste fertilization was in infields (see Bull et al. 1999), the sheer size of milpas and their distance from habitations clearly limited the use of waste and organic-matter fertilization in outfields.

Another possibility is the importation of quarried soils. Villagers today quarry soils on their hunting forays into the savanna to provide a base for specific garden plants, often flowers. These trips to the savanna can result in up to 400 kg (ca. .3–.4 m³) of savanna soils per year, but there is no limit, of course, if a household wanted to build up the soil in raised beds. This practice is in fact done on a larger scale by the modern residents of both Cobá and Chan Kom (Folan 1978; Redfield and Villa Rojas 1934:46). So far, we have tried to distinguish ancient raised beds in domestic contexts on the basis of surface survey and soil chemical tests, but we will soon make more quantifiable distinctions through excavation and further tests of soil chemistry and biomarkers (e.g., aquatic or amphibious snails) of imported savanna or estuary soils.

Organic matter rich in phosphorus might also have been composted and imported from the swamp estuary or savanna zones. This compost might then have been spread on beds. Indeed, we observed a number of rock alignments in solares that originally could have contained about 200–500 m² raised beds. We estimate that each bed would have required at least 10–20 cm deep mixture of organic matter with soil. Using these figures, each raised bed would require from 20 m³ to 100 m³, or about 10–70 metric tons, of transported organic sediment. A conservative estimate of 20–40 kg for how much a person could carry to build the beds means that it would require 250 to 1,750 trips to build the beds over a number of years and at least 25 trips per year to replenish decomposed organic matter. Since current farmers travel five or more kilometers with heavier loads of wood and soil on almost a regular basis, ancient Maya farmers could have maintained raised beds in this zone from imported organic matter. Certainly, over a period of years farmers would have had to walk farther and farther for their soil imports or actively cultivated organic matter (Beach 1998a; Palacios-Mayorga et al. 2003).

Another possibility that will continue to confound our estimates of sustainable productivity in solares for some time to come is the possible importation of periphyton (Fedick 2003; Fedick et al. 2000; Morrison and Cozatl-Manzano 2003; Palacios-Mayorga et al. 2003). These algal mats are said to be rich in phosphates, nitrogen, organic matter, and cation-exchange capacity. They are ubiquitous in seasonal and perennial wetlands and are easily transportable after drying into cakes. Beth A. Morrison and Roberto Cozatl-Manzano (2003) found small freshwater snails within higher-drier settings in the Yalahau region of northern Quintana Roo, implying the gathering of periphyton for soil enrichment, and today trucks haul large quantities of periphyton to fertilize gardens. We are nevertheless struck by the fact that milperos and dooryard gardeners outside of this zone are totally ignorant of its benefits even though they have lived intimately with it for generations.

The fairly common alignments of stones that are now flush with the ground surface in residential units could have been retaining walls for raised beds; still other raised beds might have been built of wattle or other perishable materials and are now unrecognizable. Their proximity to sascaberas and wells in solares would have greatly facilitated pot irrigation. Raised beds are often used today as nurseries to control germination and boost the vigor of seedlings and saplings before transplanting to outfields. With significant improvements in the depth of soils, raised beds could contribute significantly to cultivation. In fact, some local farmers do grow root crops in raised beds in their kitchen gardens today, but the roots mainly add variety to the diet. Based on the number of sascaberas and wells recorded so far in residential units, however, only about a quarter of all households could have had pot-irrigated nurseries.

If some form of mulching of Classic-period gardens is likely, fertilization of outfields seems much more difficult except for green manuring and fallowing. If fallow cycles in milpas were reduced much beyond ten years, sustainable cultivation would absolutely necessitate the importation and application of phosphorous and other essential nutrients. If indeed the ancient Maya were able to use all the intensification techniques discussed, then they may have been able to sustain their families based on much smaller outfields, as have, for example, rural societies in Asia. It should be clear from the figures given earlier, however, that the amount of labor seems staggering; we also have no ethnohistorical precedent and only a glimmer of archaeological evidence.

Alternative Crops Not Previously Considered

An alternative means of achieving agricultural self-sufficiency might entail xerophytic cultivars that Mesoamerican foragers and incipient cultivators used widely as basic foodstuffs—acacia, guaje, nopal, and agave (Flannery 1986). These highly productive xerophytic plants are native to northwestern Yucatan (Flores and Espejel Carvajal 1994:63–65) and are better adapted to the prevailing climatic and edaphic conditions than the more popular—and more problematic—cultivars such as maize and beans. Perhaps of some relevance is the fact that some of the earliest densely settled agricultural communities on the northern plains were located near the inner margins of the savanna (Garza Tarazona de Gonzalez and Kurjack 1981:19–20), where these highly productive forage foods are most abundant. Of course, these ecotones also would have provided early Maya farmers with access to both land and marine resources.

Acacia, or mesquite (Propsopis spp.), and guaje (Leucaena spp.) are rich in protein and carbohydrates. They are legumes, so they fix nitrogen in the soil. Both trees are found in abundance in the savanna and in secondary growth, including fallowing fields.

The genus Agave includes 136 species, of which seven or eight have been widely used for food and fiber (see Parsons and Parsons 1990). One species of Agave, also called henequen, maguey, sisal, yucca, or century plant (Agave fourcroydes), is well suited to the semi-aridity and thin soils of northwestern Yucatan and is remarkably productive in the face of the storms, droughts, and pests that ravage maize. Indeed, the commercial crop henequen virtually blanketed northwestern Yucatan in the “henequen zone” until recently; its fiber for making twine and rope has been Yucatan's most important commercial export.

Agave pulquero (as well as henequen) may have a long history in Yucatan, since its Mayan name, “Ki,” is also the term for wine (Clemency Coggins, personal communication 1997). In fact, almost every part of the agave plant—its edible leaves, hearts, inflorescences, and sap or aguamiel—can be eaten fresh, prepared in a variety of ways (e.g., in roasting pits or in roasting piles on top of the ground surface) or dehydrated for storage. The flesh is very sweet and tastes something like baked yams (Schneider et al. 1996), but because it is fibrous, it must be heavily masticated leaving a fibrous quid. The flesh and even the blossoms (Schneider et al. 1996) can be dehydrated, ground into a flour, stored for a matter of months, and then made into a bread. We have no figures on the caloric and nutritional values of roasted agave flesh. The leaves can be harvested at any time without endangering the mature plant. The plant also supports a worm or grub that is edible when roasted and salted. In the highlands, agave products are eaten most often just before harvests when stores of maize are lowest. In addition to being drunk directly as a sweet refreshing drink, the aguamiel can be fermented into pulque (and, of course, recently distilled into tequila), let to sour into vinegar, heated to convert into loaves of crystallized sugar that can then be stored or traded, and made into medicinal preparations (for ointment and/or poultice for wounds and snake bites). It is important to note that, depending on how the henequen leaves are processed for agualmiel, the fibers can be used to make a variety of cordage products afterward (Colunga García-Marín 2003).

Sustained aguamiel production can occur only by harvesting mature plants (eight to 25 years old). Therefore, most agave plants will be unproductive in any given year and will occupy a good deal of planting space. Nevertheless, Susan Evans (1992:110) estimates that in Cihuatecpan in the Mexican highlands, a 1.6 ha plot of agave intercropped with maize and consumed in the form of aguamiel produces enough calories to sustain a family of five or six persons at the subsistence level for a whole year. Moreover, David Lentz and colleagues (1996) report a Classic-period infield and/or garden that was densely planted with agave at the volcanic ash buried site of Cerén, El Salvador, and Landa (Tozzer 1941) reported its cultivation in house gardens in sixteenth-century Yucatan (also see Colunga García-Marín 2003).

The fact that there is no intensive use of Agave pulquero in Yucatan today or historically is curious, but Yucatan is not unique in this respect. Jeffrey Parsons and Mary Parsons (1990:269) note in highland Mexico that “the flesh, sap, and fiber of the maguey plant have been prepared, distributed, and consumed in a variety of ways that have either disappeared completely from the modern scene, or have been very substantially modified relative to their prehispanic analogs” (see also Evans 1992). The kinds of tools used to process agave are not terribly distinctive from ordinary domestic tools (Parsons and Parsons 1990) except, perhaps, if ceramic vessels were used to ferment pulque and roasting pits if the flesh was prepared for roasting. However, artifactual evidence of agave's use in cordage products might include larger than normal spindle whorls (Parsons and Parsons 1990). Perhaps the best chances of recognizing agave production is in microscopic wear patterns on stone tools and in plant residues. Agave leaves contain oxalate and phosphate crystals that act to store and retain water molecules, and these elongated crystals might be recognizable both in soils where the fibers were processed and on tools.

Finally, the cactus nopal (Opuntia spp.) is potentially important because it occupies a niche that most other plant life avoids: the extremely thin soils around patches of limestone pavement. Nopal has very little nutritional value, but the fruit (tuna) is relatively high in sugar, vitamin C, and many of the B vitamins (Borrego E. and Burgos V. 1986). The tuna is eaten raw or cooked. It is often made into a candy, preserve, or syrup that can be fermented into colonche, a mild narcotic drink. Tuna “cheese” was made from the ground pulp of the fruit seed. Nopal has several medicinal uses, including as an anti-inflammatory (in the form of a poultice) and in the management of non-insulin-dependent diabetes mellitus.

It seems unlikely that the ancient Maya in the Chunchucmil region used acacia, guaje, agave, and nopal as primary food staples. The questions are: How much did the the Maya rely on them? And could they have fully compensated for chronic shortfalls in maize? We have no ready answers.

Other Rock Alignments

Finally, rock alignments have just been discovered in parts of the seasonally inundated savanna to the west of Chunchucmil (Hixson 2004). They resemble rock alignments found in similar wetlands of the Holbox Fracture Zone, northern Quintana Roo (Fedick 2003; Fedick et al. 2000). According to Scott Fedick, such rock alignments may have supported some sort of subsistence activities—for example, “to control the movement of soil … [for the] cultivation of domestic crops or the encouragement of edible or economically useful wetland resources” (Fedick et al. 2000), but, like Fedick and his colleagues, we cannot specify what those activities were. The alignments may have increased the frequency and areas of ephemeral stagnant pools, which would foster the development of soils, periphyton, and reed and insect colonies, as well as attract game. It cannot be ruled out, however, that the rock alignments functioned as narrow causeways for foot traffic across the inundated savanna—to provide access to wild wetland resources such as ramon stands, wild game, palm, reeds, edible insects, as well as to small milpas located on distant tzekeles and, eventually, to the headwaters of canals leading to the coast. We remain open to the possibility that these rock alignments served both functions, just as canalized fields provide for both transportation and water management at other Maya sites. Additional studies are under way.