Climate and landscape mediate patterns of low lentil productivity in Nepal

efficiency determinants, it should be noted that the dependent variable is the one-sided error term reflecting technical inefficiency. Therefore, positive signs of the coefficient indicate tech- nical efficiency-reducing factors and negative signs reflect technical efficiency-enhancing fac- tors. Our results showed that the farm holding size, years of farming experience, farm labor availability, labor wage rate, and production specialization affect the technical efficiency of len- til production (Table 3). The results showed that more experience in lentil farming and timely availability of farm labor increases technical efficiency. Similarly, technical efficiency in lentil production decreased with increases in the on-farm wage rate. Our results are plausible because when the on-farm labor wage rate increases, pro- pensity to use labor decreases due to capital constraints, which ultimately affects crop manage- ment practices and technical efficiency. In the same vein, household labor availability is positively associated with technical efficiency. We also found strong evidence that technical efficiency increases with decreasing farm size (cultivated land). Surprisingly, lentil-specialized farms are negatively associated with technical efficiency. The histogram of the technical efficiency score estimated from the stochastic frontier model showed that on average lentil producing farms are 41% efficient (S2 Fig), indicating significant scope for improvement with the caveat that factors such as drainage class are best viewed as

intrinsic field characteristics that can only partially be modified by management. In wet win- ters, technical efficiency is higher in upland ecologies (43%) than in the lowland ecologies (39%). Results from on-farm trials Multi-locational trials were conducted across all districts in the region of interest over a five- year period (i.e. crop harvests in 2012, 2013, 2014, 2015, and 2016) to explore how crop estab- lishment practices (zero tillage), local modification of drainage (bed planting), choice of lentil variety, and integrated better bet agronomic management (for example: improved variety, alternative seeding method, timely weed management, fertilizer management etc.) affect yield and yield stability across years. None of interventions or their interactions were statistically sig- nificantly, hence we combined data across all treatments to explore location and year effects. Lentil yields varied across the years and production ecologies. Across western Nepal, wet winters were experienced in lentil season of 2012–13 and 2014–15 with poor yields attained in lowland ecologies (i.e. grain yield < 300 kg ha-1) (S2 Table). However, yields in upland sites (Dadeldhura and Surkhet) exceeded 1.0 t ha-1. In the drier winter years of 2015–16 and 2016– 17, lentil yielded more than 1.2 t ha-1 in Terai ecologies, while there was no or very poor yields in the upland sites in the mid-hills (S3 Fig). All of the results from on-farm experiments indi- cated that in wet winters, lowland areas are risky for lentil production and upland areas have

high yields, a dynamic that is reversed in dry winters. Simulating Stemphylium blight disease risk in lentil Simulation results from the Stempedia model driven by historical climatic data (1985–2014) confirms that Stemphylium disease risk varies across years but that 73% of years are likely to have significant Stemphylium blight disease outbreaks in fields where inoculum is present (Fig 2A). These results are consistent with the stochastic frontier model that has shown a negative effect of diseases with lentil production (Table 3) and Stemphylium blight could be a major rea- son. Moreover, it is important to note that local differences in drainage class and soil type influence the likelihood of disease, but these factors are not simulated in the Stempedia model and the result are best interpreted as general trends as governed by weather conditions. Results also suggest that disease severity increases with delayed crop establishment, i.e., seeding after PLOS ONE Climate and landscape mediate low lentil productivity PLOS ONE | https://doi.org/10.1371/journal.pone.0231377 April 16, 2020 12 / 19