Abstract:
Abstract: Inacurate estimation of parameters of soil water characteristic curves can cause simulation results of soil moisture models. This paper introduced firefly algorithm to solve the nonlinear fitting problem of soil water characteristic curve model (Van Genuchten) parameters. In the algorithm principle, the force of attraction and light are based on the distance between different individuals, based on which the whole population is divided into a number of subgroups at the beginning, and every sub-population gathered around local optimal value. All sub-population can optimize at the same time, improving the efficiency, and the global optimal solution is found between all these local optimal values. Compared with the particle swarm algorithm, the firefly algorithm does not use perception such as individual best position, the global optimal position to control calculation, which avoids the potential defects such as premature convergence. In addition, the algorithm dose not set speed for fireflies, which avoids the problem that the speed exceeds a threshold. But, when firefly individuals approach the optimal value, it is likely to move longer than the distance between the individual and the optimal value, therefore resulting in the jump-over of the optimal value and reaching the other side. If this situation appears repeatedly, it will affect the convergence speed and accuracy of algorithm. So in this paper we turned the fixed step length into a variant step length that changed according to the between-firefly distance, and thus the firefly algorithm had a better global optimization ability at the beginning and rapidly found the position adjacent to the optimal value region, in the end it had a good local search ability and was able to find the overall optimal value. The test samples came from test field of College of Water Conservancy and Architecture in Northeast Agricultural University(126°45'32″E、45°44'41″N), China. Black soil in Harbin area can represent the southern area of Heilongjiang. The samples were taken from four adjacent areas with a size of 10m×10m. Sampling depth was 0-20, >20-40, >40-60, >60-100, >100-140, and >140-180 cm, respectively. The relation of the soil suction and the soil moisture content was measured by a hypothermia supercentrifuge. The standard particle swarm algorithm, the standard genetic algorithm, the improved algorithm and the standard firefly algorithm were used in this paper to simulate the process of black soil and clay moisture desorption at different depths, and the relevant parameters and volumetric water content were obtained from calculation. The simulation results showed that simulation results from standard firefly algorithm and improved firefly algorithm were of high precision, in good agreement with the experimental results (the average error was less than 1%, the maximum relative error was only about 2%)by comparing calculating soil moisture characteristic curve with the experimental one. Firefly algorithm accuracy was higher than that of particle swarm algorithm and genetic algorithm, the improved firefly algorithm began to converge after about 23 iterations, the improved firefly algorithm was obviously better than the traditional ones. It was viable to use the firefly algorithm to solve this kind of nonlinear optimization problem, especially for high precision, accuracy or provide accurate basis for follow-up studies. For the surface black soil (0-20 cm) and clay in the lower layer (>20-40 cm), when the parameter range were 0.0251, 0.0496 and 0.0021, 0.0110, the maximum deviation of volumetric water content simulation results was less than 5%, indicating that the simulation precision was high. Fitting results can accurately reflect the characteristics of black soil in Harbin area, and the result of the reciprocal of air-entering value conformed to the actual situation, reflecting the characteristics of less clay and light texture in the surface layer, more clay and heavy texture in parent material horizon.