Fig. 3.

Bar graphs showing the predictive ability for 5 traits in four datasets: A BLUEQ17, B BLUEQ18, C BLUEC18, and D BLUEAll. Single-trait prediction model (ST-CV1), and multi-trait prediction mode (MT) with two schemes of cross-validation (MT-CV1 and MT-CV2. Mean Pearson’s correlations and standard error for each environment were presented for each trait. HI, harvest index; GY, grain yield in kg ha^− 1; GN, grain number m^− 2; SPI, spike partitioning index; FE, fruiting efficiency in grains g^− 1 of spike dry weight at anthesis+ 7 days; NDVI, normalized difference vegetation index; CT, canopy temperature in ^oC