Extraction of Soil Moisture Change Involved in Soybean Yield by Similarity Evaluation Encompassing Time Series Data

Our research group aims to analyze and grasp the factors for high and low yielding by using the cultivation data surrounding soybeans. Soil moisture is known to affect the growth of soybean significantly among the cultivation data,but the knowledge to directly improve the yield has not been obtained. In this study, by focusing on time series changes in soil moisture, we propose the method to discover what kind of soil moisture environment causes high or low yield. Since the behavior of soil moisture changes greatly depending on the conditions of the soil, we focus on the gas phase ratio including oxygen and estimate it . In order to grasp what time series change of the gas phase ratio affects the yield, we aim to discover cultivated spots that show similar time series change among many spots. We evaluate the similarity of time series change between spots at each growth stage because soybean requires different environments for each stage. When judging the presence or absence of similarity, it cannot be determined in advance which level or more time series change is similar. Therefore,by evaluating various levels of similarity hierarchically and comprehensively, we adopt the level of similarity that is assumed to be heavily involved in the yield. On the other hand, it may be determined by time series change in several stages, not only a single stage, whether the yield is high or low. Thus, we target the whole stage corresponding to the whole of growth to discover time series change involved in high or low yield. This method was applied to actual data of soil moisture. As a result, we discovered knowledge involved in the generally known low yield , and we also dis-covered knowledge involved in high yield based on the feature of each growth stage.