Abstract

Jason Ecker, Scott William and David Dean Colin

In four wheat trials at saline sites in coastal Bangladesh soil salinity changed continuously as the season progressed. It changed both laterally and by depth even at a sub-meter scale. The water table also moved deeper. Despite the dynamic nature of salinity and its spatial variability, the 112 individual plots in each trial maintained their general ranking for soil salinity, all changing in concert. Plot grain yield was best correlated with salinity measured as an average of 0-90 cm deep soil cores extracted within a month of sowing. Yield declined linearly and sharply at approximately 14 g/m2 (146 kg/ha) per dS/m. Standard randomisation techniques failed to distribute genotypes evenly across salinity levels in the trials. Consequently, those genotypes with most replicates fortuitously in lower salinity plots had the highest average yield. In standard screening trials lacking salinity tracking in all plots, those genotypes would have been incorrectly labelled most salt tolerant. We suggest that highly variable saline sites to be used for species and genotype screening should be mapped for salinity prior to sowing, and plots marked out to cover the range of salinity likely to produce some yield (0-20 dS/m in wheat). Selected plots should then be labelled for their relative salinity and divided into four salinity categories, high to low. Each genotype and check variety should be randomly allocated to all four salt categories at sowing. At grain harvest this design provides a four-point curve of yield versus salinity for each genotype. Assessing each genotype curve against the site curve, generated from amalgamated data of all genotypes, provides a genotype ranking of salinity tolerance and first-order tolerance benchmarking.