长白落叶松树冠光分布的动态模拟

目的消光系数（k）是模拟树冠光分布的重要指标，本研究通过对比常见的获得k值的方法，筛选最优方法，对长白落叶松人工林树冠内光合有效辐射（PAR）进行动态估计。方法（1）将实测的PAR数据按3∶1划分为拟合数据和检验数据，利用拟合数据构建k值预估模型（方法I）。（2）用拟合数据，采用人为设定不同梯度的k值估计树冠PAR，筛选最优的k值（方法II）。（3）基于叶倾角数据，采用2种不同的平均叶倾角公式（方法III-1、方法III-2），对k值进行计算。将检验数据作为独立样本对以上3种方法估计的PAR进行独立性检验。通过对比以上3种方法对树冠内PAR的估计效果，选择最优的k值计算方法，结合气象数据对PAR进行动态估计。结果根据实测PAR数据计算的树冠各轮层k值存在较大差异，总体在0.1 ~ 1.5之间，且与相对着枝深度（RDINC）呈明显的指数函数或幂函数关系。同时太阳高度角（S_a）、累积叶面积最大值（MCLA）、叶面积密度（NAD）和树冠表面积（CS）对k值的垂直变化也有明显影响。因此方法I将指数函数作为基础模型，以RDINC、S_a、CLA、NAD和CS为自变量建立了k值估计模型，模型的拟合效果较好（R2 = 0.736，RMSE = 0.124）。方法II中，当k取0.32时对PAR的估计效果最好。利用方法III计算的各轮层消光系数差异较小，总体在0.3 ~ 0.7之间。采用独立样本检验以上3种方法对PAR估计的效果，结果表明方法I对PAR的估计效果较好（平均误差ME = 2.88，平均误差绝对值MAE = 117.4，预估精度P = 91.53%），方法II对PAR的估计效果次之（ME = ? 7.2，MAE = 217.5，P = 88.12%），方法III对PAR最差（方法III-1中 ME = 121.4，MAE = 210.1，P = 55.85%；方法III-2中 ME = 226.4，MAE = 259.0，P = 42.93%）。结论k值在不同林木、不同轮层及不同的太阳高度情况下并不是一个固定值。本研究建立的k估计模型充分考虑了以上3个重要变量，符合客观实际，且对估计长白落叶松树冠PAR有良好的效果，研究结果为人工长白落叶松树冠内不同位置净光合速率的模拟提供了基础。 

Dynamic simulation of light distribution in the live crown of Larix olgensis trees

ObjectiveExtinction coefficient (k) is an important indicator to simulate the light transmission in the crown. By comparing the different methods to obtain k, this paper aims to select the optimal method to estimate the dynamical PAR in the crown of Larix olgensis trees.Method (1) The PAR data was divided into fitting data and validation data with a ratio of 3:1 and the k predicting model was developed; (2) artificially setting k with different gradients and using fitting data to estimate crown PAR then selecting the optimal k value; (3) Based on the average leaf inclination data, we calculated k with the average leaf inclination formula. The test data was used as an independent sample to conduct an independent test on the PAR estimated by the above three methods. By comparing the above three method’s estimation effect on PAR in the crown, we selected the optimal k to estimate dynamical PAR with meteorological data.ResultAccording to the measured PAR data, there was a big difference in the crown’s rotation pseudowhorls k, which ranged from 0.1 to 1.5, and showed an obvious exponential or power function relationship with RDINC. Meanwhile, solar altitude angle (S_a), max cumulative leaf area (MCLA), needle area density (NAD), and crown surface area (CS) also had significant effects on k vertical variation. Therefore, considering the exponential function as basic model, the k predicting model was established with RDINC, Sa, MCLA, NAD, and CS as independent variables, and the fitting result indicated that the k model performed well (R2 = 0.736, RMSE = 0.124). PAR was best estimated when k was 0.32. The difference of k values in each pseudowhorl calculated by the average leaf inclination distribution formula was not obvious, which ranged from 0.3 to 0.7. The perform of the above three methods on PAR estimation was tested and the results showed that the Method I performed the best (mean error ME = 2.88, mean absolute error: MAE = 117.4, precision estimation: P = 91.53%), Method II was better (ME = 2.88, MAE = 217.5, P = 88.12%), Method III was the worst (Method III-1 ME = 121.4, MAE = 210.1, P = 55.85%; Method III-2 ME = 226.4, MAE = 259.0, P = 42.93%).Conclusionk was not a constant value in the case of different trees, different pseudowhorls and different Sa. In this study, the k model was established which fully took Sa, CLA and RDINC into account. The PAR for Larix olgensis trees was well estimated based on the k model. The results will provide a scientific basis for simulating the net photosynthetic rate of live crown with different position for planted Larix olgensis trees.