干旱影响下光能利用率模型模拟常绿针叶林总初级生产力的比较

A comparison of different radiation use efficiency models in gross primary production accuracy under drought

【目的】研究干旱影响下不同光能利用率模型估算常绿针叶林总初级生产力(GPP)的能力,为精准模拟北半球常绿针叶林的GPP提供参考依据。【方法】采用MODIS数据和通量站点观测数据,在8 d尺度上对常绿针叶林的GPP进行参数化建模。首先利用皮尔逊相关系数和随机森林因子评价方法分析GPP各个驱动因子的重要性。使用降水量和潜在蒸散量(PET)构造干湿度指数,对各个站点进行干湿分类。其次在植被光合模型(VPM)的基础上,去除其中的水分参数,得到改进的植被光合模型(VPMsw)。最后定量比较各站点在不同干湿类型下,VPM模型与VPMsw模型拟合GPP的精度。【结果】(1)由驱动因子分析可知:GPP与温度、PET相关性较强,与陆表水指数(LSWI)相关性较弱,温度对GPP的重要性大于水分。(2)与VPM模型相比,VPMsw模型在干旱和半干旱站点拟合GPP的精度有所提高,均方根误差分别减少6.5%和23.4%,在半干旱地区精度提升效果更明显。【结论】干旱影响下,LSWI不能较好地反映干旱和半干旱地区常绿针叶林的水分情况,所以在干旱和半干旱地区,VPMsw模型拟合GPP的精度更高。图3表3参25。

[Objective] With an investigation of different radiation use efficiency models estimating the gross primary production(GPP) of evergreen needle-leaved forest(ENF) under the influence of drought, this paper is aimed to provide basis for accurately simulating GPP of ENF in Northern Hemisphere. [Method] MODIS data and flux site observation data at 8-day scale were used to simulate the GPP of ENF. Firstly, Pearson correlation coefficient and random forest factor evaluation method were used to analyze the correlation and importance of each driving factor of GPP while precipitation and potential evapotranspiration(PET) were used to calculate the dry-wet index to classify the dry-wet type of each site. Secondly, the VPMsw model was constructed by deleting the moisture parameter in the vegetation photosynthesis model(VPM). Finally, the accuracy of GPP estimated from VPM model and VPMsw model were quantitatively compared under different dry and wet types at each site. [Result](1) GPP had strong correlation with temperature and PET, but weak correlation with Land Surface Water Index(LSWI).(2) compared with the VPM model, the accuracy of VPMsw model in fitting GPP was improved in arid and semi-arid sites, and the root mean square error(RMSE) decreased by 6.5% and23.4%, respectively. the accuracy of GPP estimates in semi-arid sites was significantly improved by the VPMsw model. [Conclusion] VPMsw model was more accurate in simulating GPP in arid and semi-arid areas because LSWI could not well reflect the water content of the ENF in arid and semi-arid areas under the influence of drought. [Ch, 3 fig. 3 tab. 25 ref.]