FLUXCOM (RS+METEO) Global Land Carbon Fluxes using CRUNCEP climate data

Please cite both, Jung et al., 2016 and Tramontana et al., 2016, when using any of the data for publications. This data set provides 1) subfolder "raw": global land carbon fluxes (GPP, TER) on daily to annual resolution from 1980-2013 generated by 3 machine learning methods (RF, ANN, MARS) which were forced with CRUNCEPv6 meteorological data and mean seasonal cycles of several MODIS based variables (RS+METEO setup, see Tramontana et al., 2016 for details); 2) subfolder "AnomaliesClim": detrended carbon flux anomalies driven by air temperature (Tair), water availability (WAI2), shortwave radiation (Rg) as described in Jung et al., 2016. Two variants of GPP and TER refer to 2 flux partitioning methods, where '_HB' refers to Lasslop et al., 2010, and no specifier refers to Reichstein et al. 2005. Usage notes: Long-term trends and interannual variations in this data set originate exclusively from direct effects of changing climate; i.e. the climate forcing data set CRUNCEPv6 used here. Results from using other climate forcing data will also be available. Potential effects due to e.g. CO2 fertilization or vegetation greening are not accounted for. Magnitudes of interannual variations appear to be too small in the datasets, and a normalization of the anomalies is recommended when analysing interannual varibility. Analysing long-term mean TER is not recommended due to a likely bias of mean annual TER; the same hold for mean annual NEE computed from TER-GPP. The choice of flux partitioning variant is usually not critical; we would give a slight preference to Reichstein et al., 2005 variants.

使用本数据集开展学术发表时，请同时引用Jung等人2016年与Tramontana等人2016年的研究成果。本数据集包含以下两部分内容：1) 子文件夹"raw"：涵盖1980-2013年逐日至年际分辨率的全球陆地碳通量，包括总初级生产力（GPP）与生态系统呼吸（TER），由随机森林（RF）、人工神经网络（ANN）、多元自适应回归样条（MARS）这3种机器学习方法生成，以CRUNCEPv6气象数据集与基于MODIS的多变量平均季节循环作为强迫驱动数据（RS+METEO设置，详细细节参见Tramontana等人2016年的研究）；2) 子文件夹"AnomaliesClim"：包含经去趋势处理的碳通量异常值，其驱动因子为气温（Tair）、水分可利用性（WAI2）与短波辐射（Rg），相关细节参见Jung等人2016年的研究。总初级生产力与生态系统呼吸存在两种通量分区方法变体：带"_HB"后缀的对应Lasslop等人2010年提出的方法，未带后缀的对应Reichstein等人2005年提出的方法。

使用说明：本数据集的长期趋势与年际变化仅源自气候变化的直接效应，即本研究采用的CRUNCEPv6气候强迫数据集。采用其他气候强迫数据集得到的研究结果后续也将公开。本数据集未考虑二氧化碳施肥或植被绿化等潜在影响。现有数据的年际变化幅度似乎偏小，因此在分析年际变异性时，建议对异常值进行归一化处理。由于年均生态系统呼吸（TER）可能存在系统性偏差，不建议分析长期平均TER；由TER-GPP计算得到的年均净生态系统交换量（NEE）也存在同样的问题。通量分区变体的选择通常不会对结果产生显著影响，我们略微推荐使用Reichstein等人2005年的变体方案。