基于站点观测的全球1km碳通量旬尺度数据集（1999-2020）

我们提供了1-km分辨率的全球碳通量数据集（GCFD），GCFD提供从1999-2020年，十天尺度，总初级生产力（GPP）、生态系统呼吸（RECO）、生态系统净交换（NEE）三个碳通量。该数据集是以280个站点观测的碳通量作为基准，使用光合有效辐射吸收比例（FAPAR）、叶面积指数（LAI）、2米气温、向下太阳短波辐射、潜热、感热、土壤温度、土壤湿度作为协变量, 通过深度学习（CNN）方式获得。我们进行了两组实验，以获得GCFD的精度，在空间尺度上：RMSE为1.580-1.772 gC m-2 d-1，R为0.54-0.78；时间尺度上：RMSE为2.174-2.270 gC m-2 d-1，R为0.20-0.66。 GCFD基于实地观测碳通量，它可以作为现有基于模型和卫星数据集的有效补充。该数据产品可用于各种气象、生态分析和建模，尤其在需要高分辨率碳通量的应用上至关重要。有关数据集的使用及引用，请阅读说明文档。为便于使用，我们提供了两种不同分辨率的版本：30 秒（~1km）和0.1度（~9km）。

We present the Global Carbon Flux Dataset (GCFD) with a 1-km spatial resolution. GCFD covers the period from 1999 to 2020 at 10-day temporal resolution, and includes three core carbon flux variables: Gross Primary Productivity (GPP), Ecosystem Respiration (RECO), and Net Ecosystem Exchange (NEE). This dataset was developed using 280 site-observed carbon fluxes as the reference baseline, with covariates including Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Leaf Area Index (LAI), 2-meter air temperature, downward shortwave solar radiation, latent heat, sensible heat, soil temperature, and soil moisture, via a Convolutional Neural Network (CNN) deep learning approach. We conducted two sets of experiments to evaluate the accuracy of GCFD: at the spatial scale, the RMSE ranges from 1.580 to 1.772 gC m⁻² d⁻¹, with the correlation coefficient R ranging from 0.54 to 0.78; at the temporal scale, the RMSE ranges from 2.174 to 2.270 gC m⁻² d⁻¹, with R ranging from 0.20 to 0.66. GCFD is grounded on in-situ observed carbon fluxes, and can serve as a valuable complement to existing model-based and satellite-derived carbon flux datasets. This data product can be applied to various meteorological and ecological analyses and modeling efforts, and is particularly critical for applications requiring high-resolution carbon flux data. For dataset usage and citation guidelines, please refer to the accompanying documentation. To facilitate usage, we provide two versions with different spatial resolutions: 30 arc-seconds (~1 km) and 0.1 degrees (~9 km).

搜集汇总

背景与挑战

背景概述

该数据集是一个全球范围的碳通量数据集，覆盖1999年至2020年，提供十天尺度的总初级生产力、生态系统呼吸和生态系统净交换三个变量，空间分辨率高达1公里。它基于280个站点观测数据，通过深度学习模型结合多种环境协变量生成，可作为现有碳通量数据的有效补充，适用于高分辨率气象和生态建模应用。