|
Abstract
|
Despite recent development in producing chemical medicines, associated side effects have led to increased use of medicinal plants and natural compounds. Soilsalinity, especially in arid and semi-arid regions,is a serious threat to global agriculture. Nowadays, efforts have been made to find benchmarks that can effectively select salt-tolerant or salt-resistant genotypes. In this regard, the use of computer software to predict the indicescan help usfor screening the most tolerant ecotypes. The objectives of the present study were to determine the best indicators of salinity tolerance using intelligent and regression models for eighteen commercial ecotypes of mint.The seedlings were planted in plastic pots and arranged in a split factorial experiment in a randomized complete block design with four replicates. The treatments consisted of four levels of salinity (0, 2.5, 5 and 7.5 dS m-1), two levels of harvesting time, and 18 ecotypes. The plants were grown until the flowering stage and then harvested. There was a significant difference between ecotypes in terms of calculated indices at all three levels of salinity. Indicators such as TOL, MP, GMP, YSI, STI and HM showed a significant positive correlation with YS and YP at all three levels of salinity. The cluster analysis divided the ecotypes into three distinct groups based on thecalculated indices at all levels of salinity. The principal component analysis revealed that the YP, YS, TOL, MP, GMP,YSI, STI and HM weremore suitable among otherssalt stress indices.The sensitivity analysis at 2.5 dS m-1salinity level showed that the HM, STI, YSI, YI, SSI and MP indices were of higher importance than the others. At 5 dS m-1salinity level, the HM, STI, YSI, YI, GMP and MP indices showed the highest importance whereas at 7.5 dS m-1salinity level,the STI, YSI, YI, GMP and YP indices indicated the highest importance. In general, the results suggest that ANN(MLP)model (R2= 0.999) is the best model topredict at all salinity leve
|