Research on the growth rate model of Populus spp. considering environmental factors

  Objective  Forest is known as the “lung of the earth”, which is the material and spiritual basis for promoting ecosystem circulation and human survival. Populus spp. is the broadleaved tree species with the largest cultivated quantity in China. Its characteristics of fast growth, easy reproduction, strong adaptability, and short rotation period are important for keeping the balance of wood supply and demand, promoting carbon sequestration, and realizing carbon cycle. Exploring the environmental impact mechanism on its growth is of great significance to realize the efficient management of forest resources and promote the construction of ecological civilization.

  Method  In this study, a multiple regression model of Populus spp. DBH growth rate without considering or having considered environmental factors was established based on the continuous inventory data of Populus spp. in forest resources in China. Random forest (RF), gradient boosting machine algorithm (GBM) and support vector machine (SVM) were established. The optimal parameters of these algorithms were determined by the minimum RMSE. These models were evaluated by the mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R²). The level of importance of meteorological climate, terrain, stand structure and other environmental factors on the growth of Populus spp. was explored.

  Result  The DBH growth rate of Populus spp. was mainly affected by its own DBH, decreasing with the increase of DBH, and showing an obvious inverse “J” shaped trend. The model considering environmental factors had higher accuracy than that without considering. Especially, for the growth rate model of Populus spp., the R² was increased from 0.066 to 0.403. The prediction effect of the machine learning algorithm was obviously better than the regression model algorithm. Among them, the accuracy of RF algorithm was the highest, with R² reaching 0.730. The prediction result was basically the same as the actual value. The multivariate regression model, RF and GBM were basically consistent in explaining the importance of the model, while SVM had slight differences.

  Conclusion  Although the accuracy of the regression model is slightly lower than that of the machine learning algorithm, it is a white box, which can provide a basis for determining whether the DBH is abnormal in the future inventory work. The growth of Populus spp. is affected by the growth environment and closely related to the geographical location. The low altitude areas with suitable temperature and abundant precipitation and the north slope areas with gentle slope and low slope position will be more suitable for its growth. The higher the density is, the less the conducive to its growth is. In the process of Populus spp. planting, the factors of afforestation location, weather and climate should be considered first. Secondly, the rationality of stand structure, especially stand density should be considered. Finally, we should consider whether the topographic structure is suitable for afforestation projects.