Derived Optimal Linear Combination Evapotranspiration - DOLCE v2.0

DOLCE V2 is a new hybrid Evapotranspiration (ET) dataset derived by merging 11 available global ET datasets. These include BACI, FLUXCOM RS, FLUXCOM MET, ERA5-land, GLEAM v3.3a, GLEAM v3.3b,  PML-CSIRO, PLSH, MOD16, SEBS, and SRB-GEWEX. The contribution of each dataset to DOLCE V2 is based on its ability to match field observations as well as its dependence to the other parent datasets. DOLCE V2 provides time-variant estimates of its uncertainty errors, which are consistent with its agreement with field observations.
DOLCE V2 and its previous version DOLCE V1 use the same merging technique and provide monthly ET estimates and associated uncertainties over the global land. There are several improvements implemented in DOLCE V2, these include (DOLCE V2 vs DOLCE V1):
Employing a wider range of parent datasets (11 parents vs 6 parent datasets)
Incorporating more field observations to constrain the merging technique (260 sites vs 160 sites)
Finer spatial resolution (0.25° vs 0.5°).
Longer temporal coverage (1980-2018 vs 2000-2009).

Datasets employed to derive DOLCE V2: BACI
FLUXCOM RS
FLUXCOM MET
ERA5-land
GLEAM v3.3a
GLEAM v3.3b
PML-CSIRO
PLSH
MOD16
SEBS
SRB-GEWEX.
Field observations from flux tower networks including:
FLUXNET2015-tier1 and tier2
LaThuile Free Fair Use 
CarboEurope
AmeriFlux
ARM
AsiaFlux
Oak Ridge
Ozflux

The final output is a monthly, 0.25-degree dataset of terrestrial Evapotranspiration and its error estimates over 1980-2018. The dataset is provided as yearly NETCDF4 files
We used RStudio with R version 3.6.1 (2019-07-05) Copyright (C) 2019 The R Foundation for Statistical ComputingPlatform: x86_64-w64-mingw32/x64 (64-bit)

The validity of the script and the outputs were tested by:
Visual assessment of the results
Comparison with results from similar studies
Validation with in-situ observations
Using statistical metrics (detailed in the related manuscript).

DOLCE V2是一款新型混合蒸散发(Evapotranspiration, ET)数据集，由11套公开可用的全球蒸散发数据集融合构建而成。涵盖的数据集包括BACI、FLUXCOM RS、FLUXCOM MET、ERA5-land、GLEAM v3.3a、GLEAM v3.3b、PML-CSIRO、PLSH、MOD16、SEBS及SRB-GEWEX。各子数据集对DOLCE V2的贡献权重，由其与野外实地观测数据的匹配精度，以及与其他母数据集的依赖关系共同决定。DOLCE V2可提供时变的不确定性误差估算结果，该结果与数据集同野外观测数据的契合度保持一致。

DOLCE V2与其前代版本DOLCE V1采用相同的融合技术，均可提供全球陆面逐月蒸散发估算值及对应的不确定性信息。DOLCE V2相较DOLCE V1存在多项改进，具体如下：
1. 母数据集覆盖范围更广（11套相较原6套）
2. 纳入更多野外观测站点以约束融合技术（260个站点相较原160个站点）
3. 空间分辨率更精细（0.25°相较原0.5°）
4. 时间覆盖周期更长（1980-2018年相较原2000-2009年）

用于构建DOLCE V2的数据集清单如下：
BACI、FLUXCOM RS、FLUXCOM MET、ERA5-land、GLEAM v3.3a、GLEAM v3.3b、PML-CSIRO、PLSH、MOD16、SEBS、SRB-GEWEX。

本数据集所用的野外观测数据来自以下通量塔网络：
FLUXNET2015 tier1与tier2、LaThuile Free Fair Use、CarboEurope、AmeriFlux、ARM、AsiaFlux、Oak Ridge、Ozflux。

最终产出的数据集为1980-2018年间的全球陆面逐月蒸散发及其误差估算数据集，空间分辨率为0.25°，以年度NETCDF4文件格式进行分发。

本研究采用RStudio及R 3.6.1版本（发布于2019年7月5日，版权所有©2019 统计计算R基金会），运行平台为x86_64-w64-mingw32/x64（64位）。

本研究通过以下方式对脚本及输出结果的有效性进行验证：
1. 对结果开展可视化评估
2. 与同类研究的成果进行对比验证
3. 利用原位观测数据进行验证
4. 采用统计指标开展验证（详细内容参见相关研究手稿）。