A method of soil moisture content prediction is presented using fractional differential generalization of one-dimensional Richards equation for the purposes of decision support in irrigation. We propose to use water head dynamics data gathered in the beginning of growth season to perform a calibration of the model by the meta-heuristic particle swarm algorithm. Van Genuchten model is used to describe water retention curve. The cases of the known values of its coefficients along with their identification from water head dynamics data are considered. The paper presents the results of model verification for both classical Richards equation and its fractional differential generalization. For the conditions of sprinkling irrigation in the south of Ukraine, we estimate the accuracy of modelling the water head dynamics measured by the Watermark sensors and the accuracy of seasonal irrigation rate assessment. The obtained results show that the use of water transport model that take soil fractality into consideration has no effect on assessing seasonal irrigation rate but gives up to 7% increase in accuracy while predicting the time and rate of next irrigation application.

提出了一种利用一维Richards方程的分数微分泛化来预测土壤含水量的方法，目的是为灌溉提供决策支持。我们建议使用在生长季节开始时收集的水头动态数据，通过亚启发式粒子群算法对模型进行校准。Van Genuchten模型用于描述保水曲线。考虑其系数的已知值的情况以及从水头动态数据中识别它们的情况。本文介绍了经典Richards方程及其分数阶微分广义化的模型验证结果。对于乌克兰南部的喷灌条件，我们估算了由Watermark传感器测得的水头动态建模的准确性以及季节性灌溉率评估的准确性。获得的结果表明，使用考虑土壤分形性的水运模型对评估季节性灌溉速率没有影响，但是在预测下一次灌溉的时间和速率时，准确性提高了7％。