Mega-Environment Analysis of Cool-Season Turfgrass Growing Regions in NTEP Trials.

Mega-environment analysis (MEA) uses the additive main effect and multiplicative interaction (AMMI) analysis to sub-divide crop growing regions into homogeneous sub-regions according to AMMI-1 model environmental-interaction scores. MEA of yield trial data, test locations can be sub-divided along environmental interaction patterns determined by winning genotypes. MEA sub-divide growing regions into homogeneous sub-regions. As such, the same winning (genotype) roster can be targeted at the same sub-region (mega-environment, ME) thus simplifying recommendations. Currently, the National Turfgrass Evaluation Program (NTEP) sub-divide locations according to arbitrary criteria such as climatic or maintenance schedules. The 1990 perennial ryegrass (Lolium perenne L.) variety trial was analyzed using AMMI. The perennial ryegrass (PRG) trial involved 123 genotypes and 19 test locations. AMMI-1 model was suitable for MEA because AMMI-2 was most predictively accurate but interaction axis-1 (IAS-1) was almost 8 times as large as IAS-2, so IAS-2 can be ignored with little loss. NTEP locations were sub-divided according to MEA into 4 recommended sub-regions (ME) according to their interaction patterns. Sub-region 1 (ME-1) consisted of 4 test locations (Illinois, Maryland, and New Jersey), ME-2 consisted of 7 test locations (Colorado, Kansas, Kentucky, New Jersey, Oregon, Virginia, Washington), ME-3 consisted of 4 test locations (Idaho, Iowa, Maryland, Rhode Island), and ME-4 consisted of 4 test locations (Michigan, Nebraska, Ohio, Ontario). NTEP locations with in any one ME represented different climatic regions. Interaction scores (IAS-1) across all 19 locations and 4 ME were negatively correlated (-0.67) with nitrogen fertility rate as reported by NTEP cooperators. These results suggest that NTEP test locations for PRG follow an interaction pattern according to a fertility schedule but not climatic or other cultural factors. Locations from the same ME had better predictive value for other locations from the same ME compared to locations outside it’s own ME.