Fig. 3: GF models captured a signal of landrace adaptation and predicted maladaptation under post-catastrophic climates.

GF models capture current genotype-environment associations in maize landraces and were used to predict maladaptation (GF offset) under post-catastrophic climate scenarios. a Phenotypic residuals (remaining variation after accounting for experimental design) plotted against the logged GF offset of maize landrace accessions grown in common gardens. GF offset is calculated for each phenotyped accession as the Euclidean distance of the expected genotype-environment relationship at a common garden common vs the genotype-environment relationship from the accessions’ point of origin. For both phenotypes (anthesis silking interval, n = 7833; bare cob weight, n = 3175), the reported statistic is a two-sided test from a linear regression model between the logged GF offset to common gardens and the residuals for each phenotype. No adjustments were made for multiple comparisons. Points with a more negative logged GF offset indicate maize landrace accessions that are expected to be adapted to the common garden they were grown in. Yearly logged GF offset for (b) maize (c) barley (d) rice subsp. indica (e) rice subsp. japonica and (f) sorghum. The control line represents the mean logged GF offset across all years, with the shaded region representing the standard deviation of yearly means to indicate fluctuations in maladaptation (GF offset) due to normal variability in climate. For each soot scenario, lines are the averaged logged GF offset across all accessions of a species and for each year. The vertical dotted line indicates the time of soot injection into the climate models. Inlaid scatter plots are the averaged logged GF offset across all accessions of a species two years after the incident.