Simulating soil available nutrients by a new method based on WOFOST model and remote sensing assimilation

Abstract

Soil available nutrients are important for crop growth and yield accumulation. The improvement of yield and protection of the environment can be achieved by maintaining soil available nutrients at an optimal level. Currently, more than half of the available nutrients come from fertilization in a modern farm management. Appropriate fertilization can control these nutrients at an appropriate level. The precondition of fertilization optimization is acquiring the status of soil available nutrients. We proposed a new method for simulating the available nutrients to address the abovementioned issue. On the basis of the advantages of a crop model in simulating crop growth accurately and steadily, WOFOST crop model was selected as the primary model to simulate a nutrient-limited crop growth. The key parameters were calibrated through a documentary method, farm data collection, field observation, and Remote Sensing estimation prior to applying the WOFOST model. Moreover, necessary model optimizations, such as changing the structure of the WOFOST and adding a new algorithm to the soil nutrient module, were implemented. The EnKF method was selected for assimilating time-series HJ-1 A/B data into the WOFOST to realize the simulation at field and pixel scales. On the basis of the model calibration, optimization, and assimilation method construction, the simulation method for soil available nutrients was established. In this study, the theoretical basis of this new method was analyzed, and several necessary analyses were conducted to check the feasibility of the WOFOST in estimating the soil available nutrients. A lookup table method was used to realize the model simulation in a reverse order. The contents of the soil available nutrients were simulated as the output data by taking a Leaf Area Index (LAI) that was estimated from time-series HJ-CCD data as the input. This method was applied in the Shuangshan Farm in 2014. The available nitrogen (N), phosphorus (P), and potassium (K) in the maize fields were simulated. The time of the end of SAN simulation sage (ESS) can influence the simulation results; thus, we varied such time from 173 to 273 with 10 steps and repeatedly operated using the WOFOST model to obtain different simulation results of various method application times. Furthermore, we conducted several field campaigns to obtain the observation data of the soil nutrients. These data were used to analyze the precision. Results showed that the optimal time of the ESS for N, P, and K are different, and the highest R2 values are 0.68, 0.74, and 0.52, correspondingly. The average relative errors are 7.45%, 6.17%, and 9.97% respectively. This new method can reliably simulate the status of the soil available nutrients in terms of prediction accuracy, stability, and application value.