An intermediate level of disturbance with customary agricultural practices increases species diversity in Maya community forests in Belize

Indigenous societies globally manage vast tracts of tropical forest crucial for climate change mitigation. Understanding the impact of their land use decisions on biodiversity is vital for effective conservation strategies and policy support. This research focuses on the influence of customary Maya swidden agriculture on forest species diversity. The study expands upon the model proposed by R. Bliege Bird, which demonstrates how intermediate-scale anthropogenic disturbances can shape environments and contribute to the sustainability of subsistence. The hypothesis tested is that intermediate-scale swidden disturbances increase tropical canopy tree diversity, inferred from spectral analysis in community forests. The debate surrounding customary ecosystem management centers on whether conservation is an intended outcome or an unintended byproduct of Indigenous practices. Mesoamerican agroecology studies have shown that Maya communities have historically enhanced biodiversity through coupled land use strategies, ecological knowledge, and human labor. Examples include managed tree gardens (Pet Kot) and high-performance milpa agroforestry. While evidence suggests intentional forest modification, other research indicates that sustainable outcomes can emerge from interactions within complex adaptive systems, even without intentional conservation goals. A key challenge is predicting the regrowth trajectories of secondary forests following swidden, due to complex interactions between succession, landscape variables, and socio-economic factors. This complexity often leads to the misconception that swidden negatively impacts the environment. This study bridges the gap between ethnographic and ecological perspectives by using a complex adaptive systems framework to analyze the landscape-scale effects of customary Maya swidden practices on spectral diversity, utilizing remote sensing and Bayesian modeling. The Intermediate Disturbance Hypothesis (IDH) proposes that species diversity peaks at intermediate disturbance levels, neither too high nor too low. The study uses spectral diversity as a proxy for canopy tree species diversity, analyzing the non-linear relationship between swidden disturbance and spectral diversity within the context of a complex adaptive system.

The literature review explores existing research on Indigenous land management, swidden agriculture, and the intermediate disturbance hypothesis (IDH). It highlights the significance of Indigenous land rights for carbon sequestration and deforestation reduction, referencing reports such as that from the Forest Carbon Partnership Facility (REDD+). The review contrasts perspectives on the intentionality of Indigenous conservation efforts, examining studies demonstrating both conscious ecosystem management and emergent sustainability through unconscious byproducts of human action. The Mesoamerican context is examined through studies demonstrating the historical and contemporary impact of Maya agricultural practices on forest biodiversity, soil fertility, and ecosystem health. The review discusses examples of intentional forest modification, such as Pet Kot gardens and high-performance milpa systems, alongside studies illustrating emergent ecosystem enhancement in other Indigenous contexts, such as Balinese rice terraces and Aboriginal fire mosaics. It acknowledges the complexity and inherent unpredictability of tropical forest regeneration following swidden, highlighting the need for understanding the non-linear interactions between forest succession, landscape-level properties, and social factors. While acknowledging evidence for both the effectiveness and negative impacts of swidden, the review emphasizes the lack of adequate models linking household-level decisions to landscape-scale outcomes. The review sets the stage for the current study by integrating diverse perspectives to offer a more nuanced understanding of swidden’s impact on tropical forest ecosystems.

This study employed a mixed-methods approach integrating remote sensing, ecoinformatics, and landscape ecology to analyze the relationship between swidden disturbance and forest canopy spectral diversity. The study area encompassed the community forests of two Q'eqchi' Maya villages in southern Belize, totaling approximately 18,000 acres. Data collection included a drone scan using a multispectral sensor to capture high-resolution imagery of the landscape. A community-based mapping team, with extensive local knowledge, provided ground-truthing data on land use history, fallow age, and canopy tree species diversity through forest inventory plots. The drone scan provided a 5-band multispectral dataset which was processed using photogrammetry software (Pix4D) to create a composite image. Radiometric image correction was applied using data from a calibration panel and a down-welling light sensor. A semi-automated land-use classification algorithm, trained on the community mapping team's data, classified pixels into land use classes (forest, fallow fields of different ages, pastures, and recent clearings). Spectral diversity was operationalized using the concept of 'spectral species' derived from k-means clustering of multispectral reflectance values. Landscape fragmentation was quantified using landscape metrics such as edge density and patch size. The relationship between spectral diversity and landscape disturbance (measured by edge density) was assessed using Bayesian multilevel modeling. The model incorporated landscape fragmentation variables, village-specific effects, and the proportion of pasture and mature forest within each spatial sampling unit (hexagonal grid). A forest inventory survey was conducted to validate the spectral diversity index. The study used a hexagonal grid to sample the landscape across a range of spatial scales, ensuring that the scale of analysis aligned with both ecological and social processes involved in swidden agriculture. Bayesian multilevel models were used to analyze the non-linear relationship between edge density (a measure of landscape disturbance) and spectral diversity. The choice of Bayesian methodology allowed for explicit consideration of uncertainties and the incorporation of prior information to inform model parameters.

The study addressed three key research questions: (1) the association between spectral diversity and land use classes, (2) the variation in landscape fragmentation across land classes and swidden mosaics, and (3) the relationship between spectral diversity and swidden disturbance. Regarding the first question, analyses showed that spectral diversity (Shannon's diversity of spectral species) varied across land use classes, with older fallow classes exhibiting greater diversity than recently cleared areas or pastures. This is consistent with the spectral variation hypothesis which posits a correlation between spectral variance and habitat heterogeneity. For the second question, analysis of fragmentation statistics (edge density, patch area) indicated that fragmentation was lower in recently cleared areas and increased with fallow age, but decreased again in older forest classes due to patch merging. Edge density was inversely correlated with distance from the village and roads, reflecting the spatial pattern of swidden activities. The conversion to cattle pastures showed higher fragmentation than recent clearings. Addressing the third question, Bayesian multilevel models revealed a non-linear, convex relationship between edge density and spectral diversity in both study villages, supporting the Intermediate Disturbance Hypothesis (IDH). This indicated that spectral diversity peaked at intermediate levels of disturbance. Spatial predictions using the model demonstrated that higher mean spectral diversity and lower standard deviation occurred at the mean observed edge density. The strength of the IDH effect was stronger in Crique Sarco than in Graham Creek, potentially due to sampling differences. A validation analysis, correlating spectral diversity with ground-truthed biodiversity data from forest inventory plots, showed a positive correlation (Spearman's Rho = 0.77, p = 0.072), further supporting the use of spectral diversity as an effective proxy for species richness.

The findings demonstrate that customary Maya swidden practices, operating at the household level, create landscape-scale mosaics that maintain or enhance forest biodiversity, despite not being explicitly intended for conservation. This is consistent with the complex adaptive systems framework which emphasizes the emergence of system-level properties from interactions among lower-level dynamics. The findings support the IDH, indicating that an intermediate level of disturbance maximizes species diversity. This emergent sustainability is facilitated by the spatial patterning of swidden fields around villages and roads, minimizing travel costs for farmers, and by the flexible decision-making of households regarding field placement and fallow periods. The presence of diverse land classes in swidden mosaics creates more habitats and stepping-stone habitats for seed dispersers, contributing to higher beta diversity and overall richness. Mature forests in the mosaics serve as seed banks, and habitat edges support greater faunal diversity. The study acknowledges the limitations of using remote sensing to infer biodiversity but demonstrates its value in analyzing large-scale patterns and multiple successional pathways. The results highlight the importance of considering the spatial and temporal scales at which human disturbance interacts with ecological dynamics. This has important implications for understanding the resilience of Indigenous land management systems and for informing sustainable forest management strategies.

This study provides compelling evidence that customary Maya swidden agriculture, even without intentional conservation goals, can result in enhanced biodiversity through emergent processes. The findings highlight the value of integrating insights from complex adaptive systems and landscape ecology to understand the ecological consequences of Indigenous land use practices. Future research should explore the mechanisms driving the observed non-linear relationship between disturbance and diversity, including the specific roles of seed dispersal, soil fertility, and other ecosystem processes. Further research should also investigate the wider applicability of these findings to other swidden systems globally and explore the implications of these findings for policy and conservation initiatives aimed at supporting Indigenous land management and addressing climate change.

The study relied on remote sensing data from a single year, which may not fully capture inter-annual variation in vegetation or species composition. The spectral diversity index serves as a proxy for species richness and may not capture the full complexity of biodiversity, especially regarding understory vegetation or faunal diversity. While the validation analysis showed a positive correlation between spectral and ground-truthed diversity, there remains some uncertainty in the precise relationship between spectral signatures and species composition. The analysis focuses primarily on canopy tree diversity, potentially overlooking the contributions of other species or ecological components to overall biodiversity. The study area is specific to two Maya villages in Belize, and the generalizability of findings to other regions or swidden systems may vary due to differences in ecological conditions, land use practices, and socio-economic contexts.