Vichama is an archaeological site located in the modern city of Végueta, 130 km north of Lima, in the PNCC (UTM WGS84 N8779900/E212200). The site is at the right margin of the lower Huaura Valley (Huaura–Supe interbasin), at 1.5 km from the Pacific Ocean and 6.5 km northwest to the river mouth (Figure 1). The Huaura River is a permanent source of water that flows ~158 km from its origin in the Andean highlands (i.e., Sura Saca lagoon—Cordillera de Raura) to the sea and works as a natural corridor that links transversally the coast with important socioeconomic regions, such as the Mantaro basin and Eastern Andean lowlands (e.g., Alto Marañón, Alto Huallaga).

Although the lower Huaura Valley has been the object of archaeological prospections since the early 20th century (Engel, 1987; Tello, 2004), Vichama was not systematically studied until the 1970s (Aguilar, 2006; Cárdenas, 1977). Since 2007, Vichama has been intensively excavated and conserved by the Caral Project (Zona Arqueológica Caral-ZAC), sponsored by the Peruvian government and local institutions (Shady et al., 2008, 2015).

Archaeological evidence and radiocarbon dates indicate the site was occupied during two chrono-cultural periods: during the Early Formative−1 (EF-1, 1800–1500 bce), when monumental architecture flourished through six architectonic periods (P1–P6), and during the Late Intermediate Period (LIP, 1000–1300 ce), when the site became a fishing village with decreased political power in this section of the valley during a period of high marine productivity (Figure 2).

Vichama is a multicomponent complex that comprises nine sectors, including Formative monumental architecture (Vi-01) and shell middens (Vi-02, Vi-03), and architecture and cemeteries of later periods (Vi-04–Vi-09) scattered across 136 ha in the hills of Cerro Halconcillo, a rocky elevation that rises 114 m above the sea level in the middle of one desertic oasis of 305 ha (Figure 3a).

During the EF-1, Vichama was part of an emerging complex polity in the lower Huaura Valley, which would have been composed of no < 20 settlements with monumental architecture (Aguilar, 2006; Cárdenas, 1977). It is possible that Vichama represented the rise of a new sociopolitical center in the region following the decline of the Caral society. The nuclear area at that time comprised 29 buildings, including public (e.g., stepped platforms, sunken circular courts, squares, and walls with reliefs), and domestic architecture of orthogonal design spread across 25 ha (Shady et al., 2015). This area is divided into two halves, upper and lower, following the bipartite spatial division common in Andean settlements (Figure 3b). The ceremonial buildings exhibit various architectural features, including stone walls arranged in horizontal rows, architectural fills using mega-shicras, and walls adorned with high-relief decorations (Shady et al., 2008, 2015). Changes in ceremonial architecture and paraphernalia (i.e., polychrome sculptures of unfired clay and particular iconography) suggest ideological transformations.

The “visual basin” of Vichama encompasses an approximate area of 9.2 km² that included diverse ecosystems that possibly were exploited for subsistence, among them, rocky and sandy beaches, swamps, lomas (fog oases), farming fields, the river, and the sea. The coast of Huaura Valley is one of the richest areas of the Peruvian sea, which currently cumulate ~20% of the total national production of fisheries (IMARPE, 2010, p. 64). Islands and islets (e.g., Anat, Azakan) located < 1 km from the shore, potentially provided dense clam banks, marine turtles, and a variety of seabirds.

The importance of fishing is undeniable. The zooarchaeological record includes a total of 91 species (23 fish, 61 mollusks, 3 marine mammals, and 4 marine birds), recovered from primary contexts (see Supplementary material 1: S1—Marine species).

In addition, the exploitation of coastal salt-banks (i.e., Punta de Atahuanca and Salinas de Huacho, among others) to produce dry-salted fish probably was an important economic activity in the region during the Initial Formative and EF periods (Engel, 1987; Tello, 2004). Modern fishermen of Végueta employ several traditional techniques such as redes de cerco, diving, chinchorro, and arte de pinta that could be used in the past to capture the wide array of reported species (IMARPE, 2010, p. 67).

According to the last report (Yseki et al., 2024), the archaeobotanical inventory of edible plants in Vichama has yielded more than 35,000 macro-remains representing 17 taxa derived from 123 primary contexts (see Supplementary material 1: S1—Edible plants species). The botanical record includes fruits (guava, lucuma, pacay, avocado and ciruela del fraile), annuals like chili peppers (e.g., Capsicum baccatum), beans (i.e., lima bean, common bean, and jack bean), squashes, and roots and tubers (i.e., achira, sweet potato, and potato). Based on their ubiquity, it seems that fruit trees, especially guava and lucuma, as well as chili peppers, were regularly consumed and used in ceremonial activities. Similarly, the high ubiquity in offering contexts of remains of lima bean, achira, sweet potato, and potato highlights the economic importance of these plants in Vichama (Yseki et al., 2024).

The absence of Zea mays (maize) macro-remains suggests that it was probably not grown, processed, or stored at Vichama. However, a recent study reported starch grains with potential fermentation-related damage in dental calculus of individuals from Áspero (Yseki et al., 2023), with a minimal presence of maize macro-remains (Shady and Cáceda, 2008). Thus, maize consumption at Vichama cannot be ruled out based solely on macro-remains analysis, as maize may have been consumed as a beverage or flour (Yseki et al., 2024).

The marine fauna and some differences between periods suggest fluctuations in sea productive conditions linked to the upwelling of rich-nutrient marine waters. For instance, during the EF period (1800–1200 bce), there is a general trend to increasing exported productivity, especially during the EF-1 (1800–1500 bce), with some fluctuations (i.e., lower productivity between 1800 and 1600 bce, with the lowest values around 1700 bce and higher productivity between 1600 and 1500 bce) that should have influenced dietary composition, possibly with higher consumption of marine products; however, the plant record shows that agriculture was ongoing during the flourishment of monumental architecture.

These conditions, however, could be relatively different from those observed during the LIP (1000–1300 ce) when productivity reached even higher values and enhanced upwelling conditions, with its maximum at ~1000–1100 ce (Kennett and Marwan, 2015). During the LIP, the archaeological deposits are plenty of marine remains, such as fish bones and shells, fishing nets, and artifacts for marine exploitation.

While faunal and plant remains can provide insight into shifts in economic practices, stable isotope analysis is more effective for targeting the net consumption of specific food resources at an individual level. Stable carbon and nitrogen isotope analyses of collagen (δ¹³C_coll and δ¹⁵N) and bioapatite (δ¹³C_ap) extracted from human bones and teeth offer valuable insights into the diets of ancient populations (Schoeninger and Reitsema, 2023). The δ¹³C_coll is a proxy of protein sources and trophic position of consumers in the food web, as well as the photosynthetic pathway of the plants they consumed. The δ¹³C_ap instead reflects the entire dietary carbohydrates, lipids, and proteins, providing a useful means to assess the energetic constituents of the diet and photosynthetic pathways (Jim et al., 2004; Kellner and Schoeninger, 2007). In the Andean central coast, C₃ plants, which use the Calvin–Benson photosynthetic pathway, are the most prevalent and typically exhibit mean δ¹³C values of −25.7 ± 2.1‰. In contrast, among C₄, which follow the Hatch–Slack photosynthetic pathway, there are only two potentially edible domesticated plants (kiwicha—Amarantus sp., and maize—Zea mays), with typical values of approximately −11.3 ± 1.4‰. Additionally, crassulacean acid metabolism (CAM) plants have isotopic values (−10.8 ± 0.2‰) that overlap with those of C₄ (Pezo-Lanfranco and Colonese, 2024).

Usually, C₄ plant consumers show significantly higher δ¹³C values (Kellner and Schoeninger, 2007). In coastal areas, however, using δ¹³C to trace maize consumption can be challenging due to the presence of marine resources (i.e., fish and mollusks), which have overlapping δ¹³C values to C₄ plants, approximately −11.9 ± 2.1‰ (Pezo-Lanfranco and Colonese, 2024). In terms of δ¹⁵N values, herbivores tend to have values 3–7‰ higher than the plants they consume, while carnivores exhibit δ¹⁵N values 3–5‰ higher than their prey. Because marine sources have higher isotopic values compared to terrestrial sources (Minagawa and Wada, 1984; Schoeninger and DeNiro, 1984), marine fauna displays higher δ¹⁵N values than terrestrial ones, linked to trophic fractionation within complex marine food webs (Chisholm et al., 1982; DeNiro and Epstein, 1981; Hedges and Reynard, 2007; Schoeninger, 2010; Schoeninger et al., 1983). Stable isotope analysis is particularly informative when combined with Bayesian mixing models for deriving quantitative estimates of dietary composition as probability distributions (Cheung and Szpak, 2021; Fernandes et al., 2014, 2015; Phillips, 2012).

By employing a combination of stable isotope proxies from collagen (δ¹³C_coll, δ¹⁵N_coll, n = 36) and apatite (δ¹³C_ap, n = 26) of 38 human individuals from Vichama, and BSIMM estimations of the caloric contribution of the potential food sources (i.e., marine and terrestrial fauna, as well as C₃ and C₄ plants; Fernandes et al., 2014, 2015) consumed by 21 Vichama individuals (see Material and Methods for model implementation details), we aim to elucidate the primary sources of energy for two occupational periods at Vichama: EF-1 (1800–1500 bce) and the LIP (1000–1300 ce).

We tested the hypothesis that changes in marine productivity influenced the degree to which marine resources were incorporated into Andean subsistence economies (Kennett and Marwan, 2015; Wilson et al., 2024). According to paleoclimatic data, an increased marine productivity between 1800 and 1500 bce was concomitant with the flourishing of Vichama (Shady et al., 2015). However, the economic importance of plants and fish and their relative contribution to diet is unknown. It is expected that the inhabitants of Vichama, particularly during EF-1, relied more strongly on marine products as a significant part of their diet.

For comparison, the LIP (1000–1300 ce) was a time of climate cooling and decreasing SSTs, when marine productivity reached its peak (Kennett and Marwan, 2015). Because the LIP has been recognized as the period of higher development of agriculture in the Central Andes, under increasing conditions of upwelling and marine productivity, a mixed diet with a higher contribution of marine sources is expected. This reliance should be reflected in paleodietary reconstructions.

The isotopic values of 38 human individuals from Vichama were classified into the two occupations, EF period (1800–1500 bce, n = 30), and the LIP (1500–1300 ce, n = 8), used here for comparative purposes (Table 1, Supplementary material 1: S2 for individuals' contexts and bioarchaeological details and S3 for radiocarbon dates). Only a fraction of the 59 elements sampled (36 bones and 23 teeth) provided intact collagen with reliable δ¹³C_coll and δ¹⁵N values from bone (n = 21) or teeth (n = 15) according to accepted preservation criteria (DeNiro, 1985; Van Klinken, 1999). Attenuated total reflectance–Fourier transform infrared spectroscopy (ATR-FTIR) studies conducted in a subsample of 10 individuals consistently suggest that no diagenetic alteration affected the inorganic fraction of bones from Vichama contexts (France et al., 2020; see Supplementary material 1: S4 for the ATR-FTIR data). Thus, δ¹³C_ap values (n = 26) were included in our analysis to complement the paleodietary information.

The δ¹³C_coll values from both bone and tooth dentin ranged from −20.3 to −11.0 ‰, δ¹⁵N values ranged from +8.2 to + 18.0 ‰, whereas δ¹³C_ap values ranged from −14.6 to −6.8 ‰ (Supplementary material 1: S5).

Although the contribution of marine protein is evident in some male individuals (Figure 4a), the mean δ¹⁵N values are lower than expected for fishers and suggest the consumption of marine protein of a low trophic level, possibly little fishes and shellfish (with mean values of δ¹³C_coll = −12.3 ‰, σ = 0.5, and δ¹⁵N = + 10.4 ‰, σ = 1.2) over the entire sequence. No significant differences were observed for δ¹³C_coll, δ¹⁵N, and δ¹³C_ap values between females and males or between preadults and adults for both EF-1 and LIP occupations (Supplementary material 1: S5). The significant difference observed between adults and preadults of EF-1 probably resulted from the weaning effect in a couple of infants (i.e., PEACS4431 and ZAC7057) or different protein sources (i.e., ZAC7067) in some preadults.

Inter-population comparisons revealed a significant difference only for δ¹³C_ap (Mann–Whitney U = 18.50, p = 0027). In general, the δ¹³C_coll, and δ¹³C_ap indicate diets dominated by protein and energy sources from C₃ ecosystems (Figure 4b; Kellner and Schoeninger, 2007). Nevertheless, individuals from the LIP exhibited a much broader range of δ¹³C_coll and δ¹⁵N values, indicating relatively higher marine protein consumption, notably by some male individuals (e.g., ZAC7061, ZAC7059), compared to the EF-1 population.

The comparisons of protein sources of Vichama EF-1 and LIP phases with other populations from the PNCC show a trend to consume a diet intermediary between that of Áspero, the most representative site of fishermen in the Supe Valley during the Initial Formative (3000–1800 bce), based on C₃ plants and marine protein of higher trophic level, and the diet of later farmers from Quebrada Chupacigarro cemetery (middle Supe Valley, Late Formative, 500–400 bce), virtually based on plants (Pezo-Lanfranco et al., 2022). LIP diets show an overlap area greater than EP-1 with Áspero diets (Figure 5).

In addition, we estimate the proportions of dietary components estimated with BSIMMs for 21 Vichama individuals with a set of complete isotopic data (Table 2, Figure 6; see also Supplementary material 1: S6 for the model's results). These results suggest that most calories consumed came from C₃ plants (contribution ranging from 14% to 86%), followed by fish and mollusks (8% and 58%), C₄ plants (3%−19%), and terrestrial fauna (3%−11%).

Thus, most individuals from the EF-1 occupation of Vichama show values compatible with a wide farming–fishing diet, with an unexpectedly high contribution of plants to people living near the sea. The comparison of diets between EF-1 and the LIP suggests a trend of increasing marine protein consumption in the later period, fewer C₃ plants, and a discrete increase of C₄ and terrestrial protein. The Kullback–Leibler divergence is 0.141, indicating a moderate difference in the average probability distributions among individuals of EF-1 and LIP, consistent with dietary change between periods with some overlap.

Our study suggests that the two distinct occupational phases of Vichama were underpinned by different economic systems. During EF-1, dated to 1800–1500 bce, although some individuals relied significantly on marine resources for calories (>25%), subsistence was based primarily on the cultivation of C₃ plants. Plant remains, including several high-energy crops like tubers, beans, and fruits, were found mainly in funerary contexts associated with public buildings, yet our results reveal the significance of these plants at the household level beyond ritual use.

This agricultural production has likely supported Vichama's political prominence at the end of the Caral era. Cultivation seems to have been a plausible activity in Vichama during the EF-1. The agricultural area near Vichama extends approximately 970 ha, and the water sources are located < 2 km from Vichama. In the zone, the phreatic level is approximately 0.85 m depth (INRENA, 2005).

Evidence of botanical macro- and micro-remains shows the consumption of several domesticated crops in coastal settlements during the Initial Formative (3000–1800 bce; Haas et al., 2013; Shady, 2006a,b,c; Vega-Centeno, 2005; Yseki et al., 2023, 2024; Zechenter, 1988). For example, plant micro-remains from human dental calculus at Áspero, in the same region, indicate significant sweet potato and moderate potato consumption during the Initial Formative period (Yseki et al., 2023), despite their scarcity in the macro-botanical record (Shady and Cáceda, 2008).

Vichama's location is appropriate for maize cultivation due to extensive arable lands and water availability. Interestingly, the isotopic evidence obtained by this work suggests a marginal contribution of C₄ (lower than 10% in most cases) during EF-1 and agrees with the lack of maize evidence in the archaeological record. The near absence of maize in Vichama contrasts with its presence in some early sites from adjacent valleys of the PNCC (Haas et al., 2013; Shady, 2006b; Vega-Centeno, 2005; Zechenter, 1988). In Áspero, maize starch has been detected in dental calculus (Yseki et al., 2023), and in Pativilca Valley, maize pollen and starches appear as early as 2500 cal bce (Haas et al., 2013). The presence of amaranth could account for the C₄ signal detected at Vichama, although no macro-remains of this plant have been recorded at the site, unlike at Caral, located 30 km to the northeast, where amaranth has been documented in an earlier period (Shady, 2006c).

Approximately 3,000 years later, during the LIP (1300–1500 ce), when Vichama had long lost its regional political influence, the site's residents relied primarily on C₃ plants, supplemented with a greater intake of marine resources. This increased focus on marine organisms is corroborated by the presence of fishing gear, such as nets and fishhooks made from mussel shells, as well as abundant fish and shell remains, all attesting to extensive fishing activities during the LIP. While maize may have been more consumed during the LIP, isotopic equifinality complicates distinguishing between C₄ plants and marine carbon sources. Two young male individuals from the LIP (ZAC7059 and ZAC7061) exhibit values of C₄ intake of approximately 20%; however, both consumed approximately 40%−60% of calories from marine sources and show auditory exostoses, an osteological marker associated with repetitive aquatic activity in cold waters (Pezo-Lanfranco et al., 2009), suggesting that they were divers or fishers rather than maize consumers. In contrast, the only individual from EF-1 with auditory exostosis (ZAC5905, a male) shows, counterintuitively, an unexpected low proportion of marine contribution to diet and δ¹⁵N values compatible with marine resources from the lower trophic level.

The overlap between maize and marine carbon values remains a potential source of uncertainty in isotopic analyses for this region. Regardless, if maize was consumed in Vichama during either EF-1 or the LIP, it likely did not constitute a staple or significant dietary component, contributing < 30% of total dietary calories (Ambrose and Norr, 1993).

During the EF-1, terrestrial mammal consumption appears to have been occasional, with contributions from these protein sources increasing in the LIP (20% of total calories), likely due to a greater availability of camelids during this period (León, 2013). The faunal record at Vichama includes only six taxa: three bird species, commensal rodents, frogs, and camelids, the latter being the only potentially edible domesticated taxon. However, archeozoological studies of inter-period consumption are still pending. The limited availability of domesticated species in the Andes, such as camelids, guinea pigs, and dogs, along with certain wild species from the lomas and desert areas, however, suggests that terrestrial protein remained a relatively minor dietary component during the Formative period, as observed in other studies (Pezo-Lanfranco et al., 2022; Pezo-Lanfranco and Colonese, 2024). As a potential limitation of this study, the presence of little fish species can lead to mimicking the signal of wild terrestrial protein from lomas, potentially C₄ protein (Coutts et al., 2011; Cadwallader et al., 2012).

Although the archaeological record of EP-1 and the LIP support fishing activities for both periods, our results revealed that the reliance on marine products was higher during the LIP, which coincides with a time of high marine productivity, compared to EP-1. However, due to the limited LIP sample size, these interpretations should be considered provisional pending further evidence.

Finally, although some preadult individuals included in the BSIMMs could theoretically retain nitrogen isotopic signals from breastfeeding and weaning, leading to spurious interpretations, this is unlikely, as the analyzed tissues (tooth roots) correspond to older children, who typically consume an adult diet. Among the LIP individuals, two are adults and three are children or preadolescents, with sampled tissues ranging from 5.5 to 14.5 years. Previous research on weaning patterns in Andean populations places the completion of weaning between 1.5 and 3 years of age (see Pezo-Lanfranco et al., 2020 for a comprehensive review), suggesting that the isotopic values in the LIP are not influenced by the weaning process. In the case of EF-1, among two analyzed infants (PEACS 4431 and ZAC7057), only ZAC7057 shows an elevated nitrogen value, which likely indicates ongoing breastmilk consumption and could potentially mimic a high intake of marine protein in the model. All other individuals analyzed are adults.

The isotope data generated in this study suggests two distinct economic regimes that are likely mirroring changes in political organization at Vichama over time. Food production systems during the EF-1 period were likely coordinated by centralized political institutions, aimed at generating surplus through intensified agriculture. By contrast, the diversified diet observed during the LIP period reflects the absence of these centralized systems at Vichama. This shift supports the existence of a diversified economy focused on small-scale farming and fishing for household consumption, with a reduced emphasis on surplus generation. These economic systems may also have been shaped by climate oscillations and the prevailing ecological conditions.

Wet and dry alternances that fueled and inhibited marine productivity, fishing activity, and, possibly, the expansion and contraction of lomas and farming fields into the region. These phenomena are linked to complex ocean-atmosphere interactions (Salvatteci et al., 2019), including oscillations in the Southern Pacific Anticyclone, trade winds, and coastal winds with effects on marine currents (Salvatteci et al., 2016, 2019; Yseki et al., 2022).

Apart from dietary adaptation, those fluctuating climate conditions could also be reflected in sociopolitical resilience and ideological response to environmental constraints, as suggested by the presence of climate-linked deities in Vichama's iconography (e.g., frog-human deity), potentially reflecting ideological adaptations to environmental constraints (Shady et al., 2015).

Reliance on marine protein during the EF-1 period might have been limited by lower biomass and marine productivity along the Peruvian coast at that time. Between 2300 and 1300 bce, SSTs showed reduced seasonality, with cooler summers compared to present conditions (Carré et al., 2014; Loubere et al., 2013). Additionally, multidecadal fluctuations in upwelling and sea surface temperature between 1800 and 1500 bce suggest that the EF-1 period at Vichama unfolded amid highly variable oceanographic conditions (Salvatteci et al., 2016, 2019), potentially making marine resources less predictable. Interestingly, this period corresponds to a greater reliance on C₃ plants.

As argued by previous research (Caramanica et al., 2020; Contreras, 2010; Dillehay and Kolata, 2004; Lima et al., 2023; Wilson et al., 2024), these oscillations coupled with ENSO alternances, could have been produced an opportunistic intensification of certain products (fisheries/crops) when technological level, and sociopolitical conditions allow it, producing corresponding diets with verifiable marine–farming fluctuations over the time. Theoretically, under El Niño–like conditions, the concurrence of drops in marine productivity with wetter conditions (raining) along the coast and lomas flourishing, despite landslides and disasters, would lead to increased land productivity in a sort of “high-risk farming”. By comparison, under La Niña–like conditions, sea temperatures cooling, increased upwelling, and marine productivity, concomitant to a dry climate (i.e., lower rainfall and lomas contraction), would have fueled a “low-risk farming” and greater marine production, probably with more marine fauna and more maize in the record, as we can observe during the LIP.

Climate conditions during EF-1 (1800–1500 bce) at Vichama also included relatively wet periods in the highlands and dry conditions in the coastal valleys, marked by greater seasonality in river flows, with increased precipitation in the wet season and lower runoff during the dry season (Leclerc, 2023; Salvatteci et al., 2019). The development of irrigation systems could have enabled intensified agriculture during the wet season, allowing surplus production to sustain populations through the drier months. Conversely, during the LIP, cooling sea temperatures, enhanced upwelling, and increased marine productivity, coupled with drier climates marked by reduced rainfall and contraction of lomas vegetation, may have fostered a shift toward greater reliance on marine resource exploitation alongside small-scale farming (Carré et al., 2014; Leclerc, 2023; Loubere et al., 2013; Salvatteci et al., 2019).

This study offers new insights into how climate variability shaped subsistence economies in early Andean societies and expands our understanding of the trajectory of Andean civilization. In this sense, our assessment of subsistence patterns during EF-1, a period of emerging political complexity, aligns well with the idea that decreased marine productivity may have encouraged greater reliance on, and intensification of, terrestrial resources (Wilson et al., 2024). Conversely, our findings partially support the hypothesis that stable climatic conditions favor the development of agrarian societies, whereas climatic volatility contributes to political instability, decentralization, and fragmentation (Kennett and Marwan, 2015). Opportunism in the exploitation of available resources and resilience in the face of climatic challenges appear to have been key behaviors of these societies. Climate oscillations could have represented a challenging opportunity to develop adaptive buffer strategies, such as the intensification of farming or other economic activities. This may coincide with the hegemony of coastal, inland, or highland societies over different periods.

These results should be refined and tested by future research taking into consideration some limitations: (1) palaeoceanographic and climate data are fragmentary, ambiguous, and subject of interpretation and do not necessarily can be congruent with our conclusions and (2) equifinality, that is, varying combinations of food contributions that may produce the same isotopic values in a consumer, persists as a potential issue for interpretations.

This is a multiproxy stable isotope study designed to reconstruct the composition of Vichama diet in quantitative terms. Our aim is to investigate the economy of Vichama to unravel the role of marine and vegetal resources in the processes of social complexity during the EF period.

A total of 59 samples (36 bones and 23 teeth) from 38 human individuals recovered from burial contexts of Vichama were sampled for the analysis. The skeletal elements recovered for isotope analyses were mainly ribs and long bones, carpals, or cranial bones when ribs were not available. Teeth were sampled when available, and isotopic values were obtained from roots, which makes a valid direct comparison of tooth and bone samples.

Individuals' sex and age were estimated using standardized methods (Buikstra and Ubelaker, 1994). To control the potential confusing factors of breastfeeding and weaning diets, for comparisons, individuals were classified in two categories (i.e., preadults and adults) according to the tissues (i.e., teeth or bones) that originate the values (AlQahtani et al., 2010).

The chronological framework of this work is based on 8 direct [Accelerator Mass Spectrometry (AMS) radiocarbon dates from individuals' collagen, and fabrics from funerary contexts], and 12 indirect radiocarbon dates from associated material or stratigraphy (R. Shady personal communication, June 2024). Radiocarbon dates of Vichama come from the AMS facilities of three different laboratories using in-house protocols through commercial services. Individuals' dates were calibrated (bce/ce) with Calib 8.1.0 package (Stuiver and Reimer, 1993), using the Mixed Marine SoHCal 20 calibration curve (Heaton et al., 2020; Hogg et al., 2020), a local marine radiocarbon reservoir correction value (ΔR = 200 ± 20) estimated from paired samples (i.e., bone collagen and cotton fabrics) of the burial context ZAC7074. For individuals directly dated, we included the mean relative contribution of marine carbon to collagen calculated using Food Reconstruction Using Isotopic Transferred Signals (FRUITS) for each phase. Calibrated dates were rounded to 10 (see Supplementary material 1: S3). Individuals who were not directly dated were assigned to each period according to archaeological methods (i.e., stratigraphy, funerary patterns, pottery styles, and textiles' decorations).

First, samples were cleaned mechanically to remove surface contaminants. Then, samples were divided for collagen and apatite extractions. Collagen was extracted in the Institut de Ciencia i Tecnologia Ambientales of the Universitat Autònoma de Barcelona (Spain).

Samples were physically cleaned and immersed in 0.6 M HCl at 4°C for several days, for demineralization, then rinsed with ultrapure water and immersed in 0.05 M sodium hydroxide (NaOH) for 20 min at room temperature. The NaOH wash cycles were repeated as needed, typically 1 or 2 times, until no further color change occurred in the solution. Samples were rinsed three times with ultrapure water for 10 min each to ensure the complete removal of NaOH. Samples were gelatinized in 0.001 M HCl (pH3) at 80°C for 48 h. The supernatant containing collagen was filtered using polyethylene Ezee filters (9 mL, pore size 60–90 μm, Elkay Laboratories Ltd.), then frozen for at least 48 h at −20°C, and freeze-dried. Stable isotope analysis of successfully extracted samples was performed at SUERC, East Kilbride (UK) using a Delta V Advantage continuous-flow isotope ratio mass spectrometer coupled via a ConfloIV to an IsoLink elemental analyzer (Thermo Scientific, Bremen, Germany). Bone collagen (approximately 0.7 mg) was combusted in the presence of oxygen in a single reactor containing tungstic oxide and copper wires at 1,020°C to produce nitrogen (N₂) and carbon dioxide (CO₂). A magnesium perchlorate trap was used to eliminate water produced during the combustion process, and the gases were separated in a Gas chromatography (GC) column heated between 70°C and 240°C. Helium was used as a carrier gas throughout the procedure. N₂ and CO₂ entered the mass spectrometer via an open split arrangement within the ConfloIV and were analyzed against their corresponding reference gases. The International Atomic Energy Agency reference materials USGS40 (L-glutamic acid, δ¹³C Vienna Pee Dee Belemnite (V-PDB) = −26.39 ± 0.04‰, δ¹⁵N (AIR) = −4.52 ± 0.06‰) and USGS41a (L-glutamic acid, δ13C (V-PDB) = 36.55 ± 0.08‰, δ¹⁵N (AIR) = 47.55 ± 0.15‰) were used to normalize δ¹⁵N and δ¹³C values. Normalization was checked using the well-characterized Elemental Microanalysis IRMS fish gelatin standard B2215 (δ¹³C (V-PDB) = −22.92 ± 0.10‰, δ¹⁵N (AIR) = 4.26 ± 0.12‰) and/or USGS88 (marine collagen, δ¹³C (V-PDB) = −16.06 ± 0.07‰, δ¹⁵N (AIR) = 14.96 ± 0.14‰). Precision is ± 0.2‰ for δ¹⁵N and ± 0.1‰ for δ¹³C.

For bioapatite analysis, bone samples were grounded with an agate mortar and pestle and processed in the Environmental Isotope Laboratory from the University of Arizona (USA). In this case, bone samples were cleaned in distilled water with an ultrasonic cleaner and powdered using a mortar and pestle. Samples weighing ~10 mg each were soaked for 24–72 h in a 2% sodium hypochlorite solution to remove the organic material. Then, the samples were treated with 0.1 M acetic acid to remove exogenous carbonates. The samples were also allowed to react with dehydrated phosphoric acid and silver (silver removes the sulfur gases from the evolved CO₂) at 70°C in the KIEL-III Sample Preparation system. The δ¹³C values of the resultant CO₂ were measured using the Finnigan MAT252 Isotope Ratio Mass Spectrometer. The calibration of the isotopic ratios was based on repeated measurements of NBS-19 and NBS-18 with a precision of ±0.08‰ for δ¹³C (1σ).

Only individuals with valid isotope values of collagen were included in the analysis. The reliability of collagen was validated following current accepted conventional preservation criteria: C/N ratio between 2.9 and 3.6 (DeNiro, 1985) and wt%C of 26%−44% and wt%N of 11%−16% (Van Klinken, 1999). We used a subsample of 10 individuals to assess the preservation of bioapatite (i.e., mineralogy and crystallinity) with ATR-FTIR.

ATR-FTIR screening was conducted at the Laboratory of Molecular Spectroscopy, Servei de Quimica—Universitat Autònoma de Barcelona using a Bruker Alpha II FTIR compact spectrometer, coupled to an Eco-ATR. Spectra of 1 mg of bone powder were collected in absorbance (A) mode with a spectral range of 400 to 4,000 cm^−1° (), for 100 scans and a resolution of 8 cm⁻¹. Spectra generation and baseline corrections were performed with OPUS 7.8 software. To evaluate bioapatite diagenesis, we assessed five indexes (IRSF, Infrared Splitting Factor; C/P, Carbon/Phosphate; C/C, Carbon/Carbon; BPI, Type B Phosphate index; API, Type A Phophate index) relative to conventional values for well-preserved archaeological samples (France et al., 2020). See Supplementary material 1: S4 for ATR-FTIR raw data and indexes.

To identify trends of isotope values we used the “classic” scatterplots of isotopic values (i.e., δ¹³C_coll vs. δ¹⁵N; δ¹³C_coll vs. δ¹³C_ap) and rKin -Standard Ellipse Areas (Albeke, 2017). After evaluation for normal distribution with Shapiro–Wilk test for normality (α = 0.05), descriptive statistics of isotope values and comparisons by sex (i.e., males, females and undetermined) and wide age-categories (i.e., preadults and adults) were performed using Kruskal–Wallis (α =0.05) and Mann–Whitney (α = 0.05) tests with SPSS v.29 (Microsoft^®). Kernel Density analysis and graphics were performed with the function geom_density() of the ggplot2 package in R environment.

The BSIMM software FRUITS (v. 2.1.1 Beta program; Fernandes et al., 2014, 2015), was employed to estimate the proportional calorie contribution of different food sources to the diet using three dietary proxies δ¹³C_coll, δ¹⁵N and δ¹³C_ap (see Supplementary material 2 for the FRUITS model's implementation). In this research, we run individuals' estimations. The consumer data for individuals consist of their δ¹³C_col, δ¹⁵N_col, and δ¹³C_ap isotopic values. For Vichama, as a group, we use the average of isotope values and the standard error of the mean.

We charged our model with four potential food sources consumed (Food groups), with their respective composition of macro-nutrients (Food fractions: bulk, protein, and energy): terrestrial fauna, providing proteins and lipids; marine fauna, providing proteins and lipids; and C₃ and C₄ plants, providing carbohydrates and proteins. The isotopic mean values of these food groups come from previous dietary reconstructions for the Andean Central Coast (Pezo-Lanfranco et al., 2022; see Supplementary material 1: S7).

In the Central Coast of Peru, C₃ plants (including tubers, legumes, annuals, and fruits) have mean δ¹³C and δ¹⁵N values of −25.7 ± 2.1‰ and +5.3 ± 3.2‰, respectively. The C₄ plants (restricted to Zea mays in such cases) have mean δ¹³C and δ¹⁵N values of −11.3 ± 1.4‰ and +7.3 ± 3.0‰. The CAM plants have δ¹³C mean values of −10.8 ± 0.2‰ that overlap with C4, and δ¹⁵N of +3.8 ± 0.4‰. Terrestrial faunal (e.g., cervids, rodents, and camelids) show δ¹³C_coll and δ¹⁵N mean values ranging between −17.5 ± 3.2‰ and +9.0 ± 3.1‰. The δ¹³C and δ¹⁵N mean values for marine fauna (including fishes, little fishes such as anchovy and sardines, mollusks, and marine birds and mammals from modern and archaeological origin) are −11.9 ± 2.1‰ and +13.2 ± 3.0‰, respectively (Pezo-Lanfranco and Colonese, 2024). The δ¹³C values from modern fauna specimens were corrected for the “fossil fuel effect” adjusting the values by +1.5‰ (Marino and McElroy, 1991). To avoid issues with preservation (Szpak and Chiou, 2019), only values from modern plants were used to feed the model.

Isotope fractionation factors, derived from experimental studies, were established at +4.8 ± 0.5‰ between diet and collagen, +10.1 ± 0.5‰ between diet and apatite, +5.5 ± 0.5 ‰ for δ¹⁵N diet to collagen (Fernandes et al., 2014, 2015).

The weighted values of each fraction of macronutrients (lipids, carbohydrates, and proteins) followed published parameters (Fernandes et al., 2014, 2015; Pezo-Lanfranco et al., 2022; Pezo-Lanfranco and Colonese, 2024). The values of δ¹³C_ap represent the total carbon mix in the diet. Therefore, we use the same bulk value for each food group. Terrestrial and marine fauna δ¹³C bulk values were estimated as a weighted mean of lipid and protein δ¹³C values (Fernandes et al., 2014, 2015). For collagen, the carbon of the protein and energy routed to the total collagen was established at 74 ± 4% and 26%, respectively (Fernandes et al., 2012). We assumed that nitrogen was derived exclusively from proteins (100%). Lipids and carbohydrates were added to the model as “energy”. To integrate this combinatorial effect, we applied a “concentration-dependent” model (Fernandes et al., 2014, 2015).

Following previous reconstructions using FRUITS (Fernandes et al., 2014, 2015; Pezo-Lanfranco et al., 2022) the isotopic composition of each nutritional fraction (protein, carbohydrates, and lipids) was obtained from the mean values of δ¹³C_coll and δ¹⁵N_coll using the following fractionation factors: −2‰ (Δ¹³C_{protein − collagen}), −8‰ (Δ¹³C_{lípids−collagen}), and +2‰ (Δ¹⁵N_{protein − collagen}) for terrestrial mammals and −1‰ (Δ¹³C_{protein − collagen}), −7‰ (Δ¹³C_{lípids−collagen}), and +2‰ (Δ¹⁵N_{protein − collagen}) for marine animals. For plants, the offsets were −2‰ (Δ¹³C_{bulk − protein}) and +0.5‰ (Δ¹³C_{bulk − lipids}), while for the δ¹⁵N value of plant protein, the known value of δ¹⁵N recorded for the plant was assumed. The carbon weight (concentrations) of each food fraction (protein and energy) from each food group was calculated according to its macronutrient composition (Fernandes et al., 2014, 2015; Pezo-Lanfranco et al., 2022; see details in Supplementary material 2). A physiological, conservative, and acceptable range of protein consumption stipulated between 5% and 45% of the total calories was charged as “prior” (Fernandes et al., 2014). The estimates of FRUITS models reflect carbon content or equivalent calorie contributions expressed as relative contributions (adding up to 1 or 100%) with an associated 1σ uncertainty (Fernandes et al., 2015).

Finally, a Kullback–Leibler divergence test (Lyman, 2008) was performed to estimate the statistical difference between the probability distributions of the four food groups between periods (EF-1 vs. LIP). This test was executed with the Entropy package of the R environment (Supplementary material 2).