Comparison of two methods for fitting boundary line in yield gap analysis: Case study of rainfed wheat in Golestan province

In the recent year, due to increasing concerns about the future food security, researches in this field are growing globally and it requires to use proper statistic methods for estimation the amount of yield gap and determination of it’s causes. In dryland management, optimization of vegetative growth before flowering in order to maximize the utilization of available water in the soil is a fundamental principle and consequently management factors such as planting date, seeding rate and amount of fertilizer nitrogen (N) use can determine the amount of vegetative growth before flowering. Boundary line analysis is a statistical method that can quantify the response of yield to an environmental or management factor in a situation in which other parameters are variable. The objectives of this study were to introduce, use and comparison of two methods for fitting boundary line to determine the best managements and simultaneously estimate potentials and calculate wheat yield gap in three county of the Golestan province include Aghala, Gomishan and Kalaleh. One, the common method based on the least squares (LS) method and another method based on linear programming (LP) so that no point is fall above the drawn line. For this study, the required information were collected from 332 wheat field during two growing season of 2014 and 2015 in Aghala, Gomishan and Kalaleh. The investigated management factors included planting date, seeding rate and amount of nitrogen fertilizer (N) use (at the base and the road). To fit a boundary line using LS method, a curve was fitted through the maximum yield points based on the least squares method. However in LP method, a curve was fitted on the top edge of the data so that any point is not fall above the drawn line. For this purpose, the parameters of the function were changed so that the total remaining was at the lowest level, in the condition that no single negative remaining was left. Then, the optimum of each specific management factor was recognized and finally with estimating potential yield and minus the average yield in each county, yield gap was calculated for each factor and in whole. The results show the optimum of each specific management factor to reach the potential yield and optimum planting date, seeding rate and optimum nitrogen (N) use in each county were determined. The average wheat yield in drylands of Aghala, Gomishan and Kalaleh was about 2600 kg ha-1 and also potential yield was about 4700 kg ha-1 according to LS method and 4800 kg ha-1 according to LP method. So there is a yield gap about 2000-2100 kg ha-1 (44-45%). It should be noted that these results are not conclusive and it would be required to several years of testing to achieve the optimal range of each region. LP method doesn’t require dividing independent factor and determining the boundary point, and also the fitted curve done with participate of all the data. In other words, all points have their impacts on the drawn curve and this is the advantage of this method. However the lack of involvement of user experience to choice the ideal points for fitting the curve would be the weakness of this method, it seems that each of these methods in certain situations -depend on the type and distribution of data- can be useful. So user after drawing the scatter plot of data should choice the best method to fit line to the data.