Characterisation of farming system diversity in mid-hills of Nepal: Participatory validation of statistical farm typologies

In the recent decades, stratifying agricultural holdings into mutually exclusive and relatively homogeneous groups has gained momentum to tailor products, services and supporting arrangements to the structure and function of farms, farm typology, (Kaur et al., 2021). The farm types are constructed based on farm features, components, production orientation as well as farmer’s socio-economic profiles (Hoppe & MacDonald, 2013; USDA, 2021). Despite accelerated farm type development, their applicability in project targeting is challenging as robust validation is required especially when they are developed from use of unsupervised algorithms. Studies have found high variability in the number of clusters formed resulting from slight change in either number or type of variables used (Alvarez et al., 2018; Kuivanen, Michalscheck, et al., 2016). As a result, stakeholders use a few farming system attributes and engage domain experts to formulate and validate farm types (Dossa et al., 2011; van de Steeg et al., 2010). In expert-based validations, two approaches have been used: confirmatory in which the resulting typology is presented to the validation panel who compare it with their own knowledge on the local farming systems diversity (Alvarez et al., 2018) and subjective in which expert-based classification/participatory characterization is compared with statistical typology using confusion matrix (van de Steeg et al., 2010). Involving both farmers and experts in a participatory process enhance the description, categorization and interpretation (Kuivanen, Alvarez, et al., 2016) although in mid-hills of Nepal, the loci of farmer knowledge is limited due to geographical barriers which limits generalization to wider landscapes. However, multi-stakeholder approach compliments farmers experiential views with wider municipal and ward planning domains (Nyambo et al., 2019). Access to innovations, strategies, and support for improving productivity, income and resilience to shocks vary across space and social-economic profiles. Understanding these differentiations could help decision makers and service providers to tailor the development interventions and advisory. Untargeted support potentially widens skewed distribution of resources, inequality, and degradation hotspots within adjacent landscapes. Multistakeholder platforms are being established to support co-creation and enhance system synergies. According to the typology by Zepeda et al (2023a, 2023b) and earlier consultations(Jibesh Kumar et al., 2023; Neupane et al., 2024), farming systems can be grouped into four distinct types in mid-higher altitude of Khotang and into 5 farm types in Surkhet. The farm-types depict the varying levels of resource endowments from farms that depend on remittances to those with strong agricultural income stream, located in varied agro-ecologies with different production systems that include millets in high-altitude, in remote areas with maize and livestock, marginal maize landscapes, to small intensive rice-wheat systems, supported by innovations for system reconfiguration such as emergence of home gardens, irrigation, forage, feminization of labour, mechanization and markets. Against this backdrop, this study was conducted as a confirmatory evaluation of the typologies proposed by Zepeda et al (2023a, 2023b) to explore which farm and household attributes farmers use to differentiate themselves. The additional objectives of the study were to assess if the current strategies used in targeting by stakeholders correspond to the typology and to explore potential entry points for innovation setting the ex-ante probability of adoption given related risks and resource envelop.