Abstract Miscanthus is a perennial C4 crop whose lignocellulose can be used as an alternative to the production of biosourced material. Miscanthus x giganteus (M. x giganteus) has demonstrated high maximum yields but also high yield variability across farmers’ fields. Miscanthus sinensis (M. sinensis) can be an alternative to M. x giganteus because it is considered to be more tolerant to water stress and to produce more stable yields. This study aimed to identify the main factors explaining the variability of yields across site-years for M. x giganteus and M. sinensis. A multi-local and multi-year trial network was set up in France (Ile de France and Center regions). Four treatments were established on seven sites, from spring 2013 to winter 2019: at each site, two treatments of M. x giganteus (a treatment from rhizome and a treatment from rhizome-derived plantlets) and two treatments of M. sinensis (a treatment from seed-derived plantlets established in single density and a treatment from seed-derived plantlets established in double density). We experienced 5 years of harvest because miscanthus was not harvested in 2014. First, we characterized yield variations across site-years for both genotypes. Second, we defined and calculated a set of indicators (e.g., water stress indicator, sum of degree-days of the previous year, number of frost days) that could affect miscanthus yields. Finally, we performed a mixed model with re-sampling to identify the main indicators that explained yield variability for each genotype specifically. Results showed that water stress and crop age mainly explained yield variability for both genotypes. M. sinensis yields were also affected by the sum of degree-days of the previous year of growth. Hence, genotype choice must take into account environmental characteristics. M. sinensis could indeed achieve higher and more stable yields than those of M. x giganteus in shallow sandy soils or in locations with a higher risk of low rainfall.