The shift from taro (Colocasia esculenta) to rice (Oryza sativa) cultivation and consumption between 1570 CE and 1800 CE was a remarkable transformation in Ifugao’s socio-ecology. To convey the scope of this transformation, this section offers a brief history of Ifugao province and OKV during the Spanish colonial period. It then describes Ifugao society circa 1570 and 1800, focusing on the emergence of Ifugao’s distinctive agroforestry complex, which emerged as a response to both Ifugao’s geography and the Spanish occupation of the surrounding lowlands. We refer to this as the muyong-uma-payo complex (see Table 1).

Contemporary Ifugao province, which approximates the Ifugao cultural region, is enclosed by the Magat River to its south and the Cordillera Gran’s steepest mountains to its north and west (Wernstedt and Spencer, 1967; Conklin, 1980; see Figure 1). As a result, the province’s topography is defined by deep river valleys where rice cannot grow except in constructed environments like terraces. Terraced rice is able to grow up to 1,600 m above sea level (m ASL), and, not coincidentally, the average elevation of the province’s inhabited regions ranges from 1,000 to 1,500 m ASL (Conklin, 1980). The region above 1500 m ASL comprises broadleaved montane forest interdigitated with Benguet pine- (Pinus kesiya Royle ex Gordon) dominated coniferous forests on steep slopes and ridges as well as in frequently burnt zones (Kowal, 1966; Jain, 1992; Olson et al., 2001; Rabena et al., 2015; Lamoreax, 2021). We refer to this holistically as coniferous montane forest for brevity. Land use in Ifugao settlements—the payo, but also the uma and muyong—take full advantage of this complex geography and the transitional zone (ecotone) from mid-to high-elevation (Figure 2).

Spanish colonizers became aware of the Ifugao region after establishing control over lowland Luzon in a series of campaigns between 1572 and 1574 CE. Cash-strapped colonial administrators and Spain’s kings were attracted by rumors of gold mines in the western and southern Cordillera, but Ifugao lacked these precious metals (Blair and Robertson, vol. 20, 1909; Morga, 1971; Scott, 1974). As a result, regular contact between the Spanish and Ifugao peoples did not take place until the mid-eighteenth century with the establishment of Ituy and Paniqui provinces (Antolin and Scott, 1970; see Figure 1). Works from that period by friars Antolin in 1789 (see ibid) and Molano in 1801 (cited in Scott, 1974) offer some of the earliest written descriptions of Ifugao. Molano’s account is of special interest since he was the first Westerner to visit OKV, the largest settlement of the Tuwali-Ifugao, and to describe Ifugao’s unique rice terraces (see Figure 2).

Although Molano may not have been aware, OKV held mythic significance to the Tuwali-Ifugao and was an unusually large settlement with fertile land. Oral tradition indicates OKV was the birthplace of rice terrace farming and its associated rituals (Maher, 1984). Writing of the village in the late 20th century, Maher said, “What remains of it can be found there in the Barangay Bullaga surrounded by pond fields on the flood plain of the Ibulao River. The surface is flat, requiring no terracing, and water is abundant and accessible” (p. 119). This corresponds with Antolin’s 1789 descriptions of “Kiangan (Kiyyangan)” and Tuwali-Ifugao lands as fertile compared to other parts of the Cordillera (Antolin and Scott, 1970, 237). However, the village’s size also made it a target. In 1832 CE, a Spanish expedition led by Commander Galvey sacked the village and much of the region. By the 1860s, the village was abandoned (Scott, 1974; Scott, 1982).

OKV’s relative prominence in Spanish documents and its mythological importance to the Ifugao make it an ideal site for land use modelling. Multiple Spanish interactions with the village mean it is relatively well-documented compared to other Ifugao settlements. Furthermore, because of the village’s mythological importance, it has been thoroughly investigated by the Ifugao Archaeological Project (IAP, see ¹ ) and is one of the best-researched sites in the province. It therefore is one of the few sites in Ifugao province where historic land use modelling may be attempted, although the village’s exceptional size and fertility mean it is not necessarily a template for the whole province. Nonetheless it is the supposed birthplace of the terrace-based rice cultivation that radically transformed Ifugao between 1570 (pre-Spanish) and 1800 (pre-conquest), and it is therefore an important site for understanding Ifugao agriculture.

Based on Molano’s written accounts (see Scott, 1974), the pre-conquest socio-ecology of OKV was defined by the payo. The payo, though, were only the most visible aspect of a highly integrated agroforestry system, the muyong-uma-payo complex (see Figure 2). The payo were also a basis for social differentiation among the Ifugao (see Table 1). They were constructed in the absence of a central administrative state, but gave rise to a “petty plutocracy” (Scott, 1982, p. 135) where communal labor maintained and expanded terraces owned by the kadangyan, whose regular consumption of difficult-to-grow rice was a primary signifier of their elite status. Complementing the payo were the muyong, carefully managed woodlots that contained as many as 200 plant species that were used for food, construction, fuel, and medicine (Sajor, 1999). Located uphill of terraces, muyong supported rice cultivation by hindering runoff and erosion, simultaneously protecting the terraces and ensuring regular access to water. Finally, kadangyan who did not own many terraces as well as Ifugao who did not own terraces, the nawotwot, consumed root crops such as iuktu (ubi, purple yams, Dioscorea alata) or, more recently, gattuk (camote, sweet potatoes, Ipomoea batatas) grown in uma (swidden fields). Uma were typically located near muyong uphill of payo on arable land unsuitable for rice cultivation (Conklin, 1980; Sajor, 1999).

Less is known about the pre-Spanish socio-ecology of Ifugao. Palaeobotanical evidence suggests rice was entirely absent, that taro and root crops were OKV’s primary starches before 1570, and that these crops were grown in less-extensive terraces and swidden fields (Acabado S. B. 2012; Acabado S. 2012; Horrocks et al., 2018). Studies employing radiocarbon dating indicate terracing substantially increased in the 17th and early 18th centuries, after the Spanish arrived in the Philippines (Maher, 1973; Conklin, 1980; Acabado et al., 2019). Based on research performed by the IAP in OKV, Acabado theorizes that increased terracing was attributable to population growth partially caused by an influx of migrants from the lowlands where rice cultivation was practiced. This population increase catalyzed both the rapid expansion of pre-existing terraces and the adoption of rice, a more productive staple, amongst social elites (Bray, 1994; Acabado S. B. 2012; Acabado S. 2012; Eusebio et al., 2015; Acabado, 2017). The shift to wet rice cultivation, Acabado asserts, provided the basis for pronounced social differentiation amongst the Ifugao, helped them resist multiple Spanish incursions and ultimately helped create a pericolonial society (Acabado, 2017; Acabado, 2018).

We use a historical land use model to quantify the amount and type of land used by the residents of OKV. The model, first conceptualized by Gregg (1988) and elaborated in subsequent studies (Kay and Kaplan, 2015; Weiberg et al., 2019), posits that land use is a function of the properties of the physical environment and the socio-cultural characteristics of the people who use the land. The former includes topography, climate, soils, and vegetation and can be reconstructed for the past based on modern survey and paleoenvironmental proxies. The latter consists of the most significant components of a population’s diet and their methods of agriculture, agroforestry, animal husbandry, hunting and gathering, pottery production, and metallurgy. Reconstructing the socio-cultural characteristics of past societies draws upon published archaeological, ethnographic, and anthropological research.

Hughes et al. (2018) describe the past land use modeling process in detail, and term the socio-cultural characteristics “livelihoods”. These livelihoods describe the human inputs to food production that, along with the properties of the physical environment, are used to determine the yield of the principle food crops and animal products. With an initial estimate of the size of the human population, the land use model is used to approximate the amount of land required for a settlement to subsist for a single year.

Here, our model quantifies land use for an imagined year where no significant demographic change, famine, or abnormal disruptions transpire. Since natural hazards are frequent in the Philippines, our model assumes moderate damage from typhoons and earthquakes, and therefore expanded land use associated with repairs, was part of a typical year (Bankoff, 2009). However, since statistics on all but the most exceptional historic typhoons and earthquakes are rare, particularly in the Cordillera region, we are forced to make generalized assumptions about the damage they caused every year (see Bankoff, 2003). It is also essential to note that our model attempts to quantify land use at specific points in time, but does not make any assumption on how or when change transpired. As a result, our models of OKV at two timepoints only identify changes that happened in intervening years, not their cause or tempo.

Following Kay and Kaplan (2015), we recognize that different forms of land use lead to different transformations of the physical environment. For example, a village or garden would represent an obvious anthropogenic modification of the natural landscape, while a forest in which hunting and gathering takes place might only be appreciated by someone intimately familiar with the local environment. Thus, land use leads to a hierarchy of impacts on the landscape. For Ifugao, we pre-define the types of land use calculated by the model and their relative intensities (Table 1).

The land use model is coded in an Excel spreadsheet and contains user-definable parameters describing both livelihood and properties of the physical environment that can be adjusted based on uncertainties in diet, inputs, climate, population size, etc. The model output consists of a table of the area in hectares required for each type of land use.

In order to aid in the rapid interpretation of the model results and facilitate comparison between outputs using varying underlying assumptions or for different periods, we use the circle diagram (Kay and Kaplan, 2015; Hughes et al., 2018). A circle diagram presents the land use model output as a series of colored concentric rings, where each ring represents a specific land use at a certain hierarchical level (Table 1). Where two or more land uses have the same relative intensity, both are shown as distinct sectors in the ring at that level. The integrated area of each ring/sector represents the area required for that particular land use. We use a custom script for the Generic Mapping Tools software (Wessel et al., 2019) to take the tabular output from the land use model and plot the circle diagrams. The full plotting code and model output used in this paper are published on Zenodo (Kaplan, 2022).

Our models for OKV in 1800 rely on 20th century ethnographic research supplemented by palaeoecological and archaeological research in Ifugao, especially the on-site investigations at the OKV by the IAP. We assume 20th century ethnographic research is useable for the 1800 timepoint based on archaeological evidence and Ifugao oral tradition, which indicate the terraces and their associated social institutions are at least ten generations old (Lambrecht, 1967; Maher, 1973; Conklin, 1980; Acabado, 2017). We acknowledge that relying on ethnographic evidence assembled in the previous century is potentially problematic as Ifugao society was and is dynamic. However, the lack of regular historic documentation of Ifugao forces us to make these assumptions. For the 1570 timepoint, for which no written historical documents exist, we rely exclusively on archaeological and palaeoecological evidence and assert that OKV in 1570 CE grew taro as its primary subsistence crop to support a smaller population. As a result, both our pre-Spanish and pre-conquest models are built upon several informed assumptions, which are a source of uncertainty in our models. To explore how these uncertainties can produce significant variations in our model results, we also run a series of simulations (or “sensitivity tests”) covering a range of input values.

The remainder of this Methodology section is dedicated to describing the information we use to assemble our model calculations with a focus on our template or initial models. These models represent what we believe to be the probable land use of OKV at our two timepoints. This section is organized according to the three broad categories of data used in land use modelling: population, “livelihood,” and yields and nutrition. Throughout, we summarize the data we gathered and the assumptions we made. When necessary, more detailed descriptions of our data are included in the Supplementary Material. We also allude to points of uncertainty in our templates and how we test the potential error introduced by these uncertainties. These are discussed in detail in the Supplementary Material, where we demonstrate they do not alter our primary conclusion: that intensification, particularly expanded terracing, and the initiation of widespread rice cultivation allowed OKV to have lower per capita land use in 1800 than in 1570, even though the village’s population in 1800 was double what it was in 1570.

Our template diagram for OKV in 1800 CE is presented in Figure 3. We assume the village contained 284 houses and 1,136 individuals, the maximum population we allowed for in Population and Boble (Settled Area) at OKV. As per our Methodology (Section 3.3), we expect that iuktu (purple yam) and gattuk (sweet potato) comprised equal portions of diets at OKV; that a small herd of carabao (∼22 animals) were present in the village; and that goats were absent from the village. Lastly, for the sake of clarity, all land use numbers presented here, excepting the Settlement, are approximations. For more precise numbers, see the datasheets included in the Supplementary Material.

We calculate that a village of this size, operating within the constraints listed above, would require ∼2,113 ha of land to subsist for a single year, with the majority of this land being hunting land. Hunting would require 1,390 ha, while all other forms of land use would require ∼723 ha of land. Assuming that the historic yield of rice was 1650 kg ha⁻¹; that historic taro’s yield was 2000 kg ha⁻¹; and that the village produced 30% more rice than required for subsistence, the village would have required ∼112 ha for payo (terraces). Of this terraced land, ∼15 ha would have been required to grow taro, despite it only comprising 2.2% of annual diets in our template model (compared to rice, which represented 32% of annual diets).

Furthermore, we expect houses and hamlets were scattered among these terraces (see section 3.3.1), so the 5.55 ha of land comprising the settlement circle was not consolidated around a single nucleus. This lowered population density and maximized individual access to the terrace paths and bunding, so the whole village’s kalabasa (squash) and other garden vegetables could have been grown in personal gardens or along terrace bunding and pathways (Conklin, 1980). In this way, kalabasa or garden vegetables were fully integrated into other forms of land use.

In addition to the payo, our model asserts that OKV required ∼120 ha of active uma (swidden fields) to grow iuktu and gattuk each year, as well as ∼240 ha of fallow land. Additionally, as discussed in section 3.1, supporting the payo and uma were the muyong, the indispensable managed forests. We calculate that OKV would have required ∼6 ha of muyong for fruits and another ∼85 ha for fuel and other essential resources. Therefore, the village as a whole would have required ∼91 ha of muyong in total to support 1,136 individuals circa 1800.

Based on the community of 1,136 individuals’ caloric needs as well as the physical requirements and breeding habits of pigs and carabao, we expect OKV would have contained ∼391 pigs and ∼22 carabao. The carabao, as grazing animals, would have required ∼93 ha of pastureland (Conklin, 1980). We expect pigs, as foraging omnivores, were primarily fed agricultural waste—failed tubers and the green material from various crops—and human waste products (Gregg, 1988; Conklin, 2002). The 10% of their diet derived from forest browse would have required ∼61 ha annually. Conklin’s (1980) research on ritual sacrifice suggests these herds would have been too small to meet OKV’s ceremonial obligations, and therefore commerce with the lowlands to obtain additional pigs and carabao was essential, which agrees with all ethnohistorical evidence (Scott, 1966; Scott, 1974; Scott, 1982).

By using Acabado’s pericolonialism theory to argue that wide-spread terracing transpired only after the Spanish arrived in Luzon, we can create a feasible model for OKV in 1570 CE. This model describes a configuration of pre-Spanish land use in Ifugao that can be tested in future research. Our template diagram for OKV in 1570 is presented in Figure 3. This diagram assumes the population was 568, and the settlement circle was ∼2.7 ha, half the size of the settlement in 1800, and no rice was grown in the village as per the pericolonialism hypothesis (see also Horrocks et al., 2018). Instead, terraces were used to grow a portion of OKV’s taro. We assume taro constituted 32% of annual diets, while yams comprised 44.3% of annual diets (sweet potatoes not having been introduced yet constituted 0% of diets). We expect 12% of taro was grown in terraces while the remaining taro and all yams were grown in swidden fields. This would require ∼177 ha of active fields, ∼354 ha of fallow fields, and ∼13.3 ha of terraces (meaning the agricultural and residential terraces together amounted to ∼16 ha of land). Additionally, our model calculations predict half as much highly managed forest would have been needed for a population of 568, so OKV in 1570 would have required ∼46 ha of muyong (3.02 ha for fruits, 42.60 ha for fuel). And lastly, our model expects ∼30 ha of unmanaged forests were used for pigs to browse.

Zooarchaeological evidence is inconclusive on whether OKV in 1570 possessed or consumed carabao. For this model, as per Acabado’s hypothesis (2017), we assume carabao were not consumed or raised in OKV in 1570. We do expect OKV raised and consumed domesticated chickens and pigs as well as hunted game in similar proportions to OKV in 1800. However, because the Ifugao region’s population density would have been lower (half what it was in 1800, based on our models’ existing assumptions), the amount of land available for hunting would have been greater in 1570 than in 1800.

In total, our model concludes OKV circa 1570 would have required ∼2,013 ha of land. Of this land, 623 ha of land would have been dedicated to all activities except hunting. This is in contrast to OKV in 1800, which required ∼2,113 ha of land, of which only ∼723 ha were dedicated to all activities aside from hunting. Therefore, our models suggest increased terracing and the switch from taro to rice, as well as the introduction of sweet potato, permitted OKV’s population to double without greatly increasing the amount of land devoted to starch production. Further, even the addition of grazing animals like carabao by 1800 does not change or challenge this conclusion. This minimization of high intensity agricultural activity undoubtedly helped OKV preserve wild deer and pigs’ habitats and therefore continue to practice widespread hunting. This, and the highly sustainable muyong-uma-payo system of integrated agriculture all indicate historic Ifugao society was premised on a stable pattern of land use that, when bolstered by commerce with the lowlands, permitted large populations to survive and maintain independence in the Cordillera.

The points of uncertainty in our model, discussed in the methodology, do not challenge this conclusion. In the Supplementary Material, we model scenarios for the 1800 CE timepoint with the potential to change our conclusion. We do not model for 1570 because less palaeoecological data is available for this timepoint, and our model for it is primarily based upon the pericolonialism hypothesis. Our scenarios for the 1800 timepoint include: varying the population of OKV; varying the daily caloric intake of individuals at OKV; varying the amount of gattuk and iuktu grown at OKV; removing the carabao herd at OKV; adding a goat herd to OKV; modelling a failed rice harvest at OKV caused by a strong typhoon or similar natural hazard; and modelling how hunting ranges could be affected by Ifugao’s historic population density. These scenarios confirm the amount of land used for all activities except hunting varied from 543 ha at minimum to ∼998 ha at maximum. All confirm intensive land use would not have increased proportionally with population, suggesting Ifugao land use in 1800 was markedly more consolidated than in 1570; was remarkably efficient; and did not rely on extensive deforestation but instead maintained muyong as well as public forests and hunting grounds.

The land used to feed a family of kadangyan was distinct from the land used to feed a family of nawotwot, and the land use model calculations can reflect this (Hughes et al., 2018). In Figure 4, we present two circle diagrams depicting the land use associated with two families of four individuals. Panel A depicts kadangyan land use and panel B depicts nawotwot land use. For this diagram, we gave each family an idealized diet, representing the wealthiest kadangyan family possible and the poorest nawotwot family possible. The kadangyan in this diagram are imagined to have consumed only rice and taro as starches, while the nawotwot consumed only yams and sweet potatoes as starches. Both imagined families, however, are presumed to consume the same proportion of starches annually (∼76.2% of their total diet). Additionally, the kadangyan in our scenario consumed carabao and a larger proportion of domesticated pigs than the nawotwot, who would have owned and sacrificed fewer pigs and no carabao (see section 3.3.5).

Our models indicate that a kadangyan family of four, given these parameters, would require 6.62 ha of land to subsist for a year. That land would be divided as follows: 0.02 ha for living space, 0.84 ha for terraces, 0.33 ha for pastureland, 0.32 ha for muyong, 0.21 ha of forest for pigs to browse, and 4.89 ha for hunting. By contrast, the nawotwot family would require the same amount of land for hunting, muyong, and living space, but no land for terraces or pasture for carabao. Instead, they would require 0.77 ha of active uma, 1.55 ha of fallow uma, and 0.13 ha forest for pigs to browse. Altogether then, the nawotwot family would need 2.79 ha of actively modified land per year to subsist, while the kadangyan family of the same size would require 1.73 ha of land.

The diagrams presented in Figure 4 help explain how land use in the 1800 timepoint did not double from 1570, despite the population doubling. Payo, a hallmark of the kadangyan class, use less land than uma. In a village like OKV, which evidence suggests was more suitable to rice agriculture than most of Ifugao (Barton, 1922), this means a high proportion of the population could have utilized payo, and their consumption of rice would have helped decrease land use even as the population grew. Nonetheless, it should be noted that a kadangyan family of four could not exist in a vacuum. While terrace construction does not require generations of laborers as once supposed (Acabado et al., 2019), building and maintaining them does take a village. As Conklin (1980) observed, a hectare of highland terrace requires 630 human days of labor per year to maintain, while a swidden field of the same size only requires 250 days. Therefore, without a large population, the Ifugao terraces would not be viable. The same principal applies to hunting for both kadangyan and nawotwot. Hunting was not undertaken by family units but instead was a community endeavor. Again, the hypothetical families we model here did not exist in a vacuum, but were instead part of a community that enabled them to enact this specific land use pattern.