Accurate estimation of gross primary production (GPP), the amount of carbon absorbed by plants via photosynthesis, is of great importance for understanding ecosystem functions, carbon cycling, and climate-carbon feedbacks. Remote sensing has been widely used to quantify GPP at regional to global scales. However, polar-orbiting satellites (e.g., Landsat, Sentinel, Terra, Aqua, Suomi NPP, JPSS, OCO-2) lack the capability to examine the diurnal cycles of GPP because they observe the Earth's surface at the same time of day. The Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS), launched in June 2018, observes the land surface temperature (LST) at different times of day with high spatial resolution (70 m × 70 m) from the International Space Station (ISS). Here, we made use of ECOSTRESS data to predict instantaneous GPP with high spatial resolution for different times of day using a data-driven approach based on machine learning. The predictive GPP model used instantaneous ECOSTRESS LST observations along with the daily enhanced vegetation index (EVI) from the Moderate Resolution Imaging Spectroradiometer (MODIS), land cover type from the National Land Cover Database (NCLD), and instantaneous meteorological data from the ERA5 reanalysis dataset. Our model estimated instantaneous GPP across 56 flux tower sites fairly well (R² = 0.88, Root Mean Squared Error (RMSE) = 2.42 μmol CO₂ m⁻² s⁻¹). The instantaneous GPP estimates driven by ECOSTRESS LST captured the diurnal variations of tower GPP for different biomes. We then produced multiple high resolution ECOSTRESS GPP maps for the central and northern California. We found distinct changes in GPP at different times of day (e.g., higher in late morning, peak around noon, approaching zero at dusk), and clear differences in productivity across landscapes (e.g., savannas, croplands, grasslands, and forests) for different times of day. ECOSTRESS GPP also captured the seasonal variations in the diurnal cycling of photosynthesis. This study demonstrates the feasibility of using ECOSTRESS data for producing instantaneous GPP (i.e., GPP for the acquisition time of the ECOSTRESS data) for different times of day. The ECOSTRESS GPP can shed light on how plant photosynthesis and water use vary over the course of the diurnal cycle and inform agricultural management and future improvement of terrestrial biosphere/land surface models.

准确估算初级生产总值（GPP），即植物通过光合作用吸收的碳量，对于理解生态系统功能，碳循环和气候碳反馈非常重要。遥感已被广泛用于量化区域到全球范围内的GPP。但是，极地轨道卫星（例如Landsat，Sentinel，Terra，Aqua，Suomi NPP，JPSS，OCO-2）缺乏检查GPP昼夜周期的能力，因为它们在一天的同一时间观测地球表面。2018年6月启动的空间站生态系统星载热辐射计实验（ECOSTRESS），从国际空间站（ISS）以高空间分辨率（70 m×70 m）观察一天中不同时间的地表温度（LST）。 。这里，我们使用ECOSTRESS数据通过基于机器学习的数据驱动方法来预测一天中不同时间的具有高空间分辨率的瞬时GPP。预测GPP模型使用了瞬时ECOSTRESS LST观测值，以及来自中分辨率成像光谱仪（MODIS）的每日增强植被指数（EVI），来自美国国家土地覆被数据库（NCLD）的土地覆被类型以及来自ERA5重新分析的瞬时气象数据数据集。我们的模型相当不错地估计了56个流量塔站点上的瞬时GPP（R 预测性GPP模型使用了瞬时ECOSTRESS LST观测值以及中分辨率成像光谱仪（MODIS）的每日增强植被指数（EVI），国家土地覆被数据库（NCLD）的土地覆被类型以及ERA5重新分析的瞬时气象数据数据集。我们的模型相当不错地估计了56个流量塔站点上的瞬时GPP（R 预测GPP模型使用了瞬时ECOSTRESS LST观测值，以及来自中分辨率成像光谱仪（MODIS）的每日增强植被指数（EVI），来自美国国家土地覆被数据库（NCLD）的土地覆被类型以及来自ERA5重新分析的瞬时气象数据数据集。我们的模型相当不错地估计了56个流量塔站点上的瞬时GPP（R²  = 0.88，均方根误差（RMSE）= 2.42  μ摩尔CO ₂米^-2 小号^-1）。由ECOSTRESS LST驱动的瞬时GPP估计值捕获了不同生物群落塔GPP的日变化。然后，我们为加利福尼亚中部和北部制作了多个高分辨率的ECOSTRESS GPP地图。我们发现GPP在一天中的不同时间发生了明显变化（例如，早晨更高，中午达到峰值，黄昏时接近零），并且在不同景观（例如热带稀树草原，农田，草地和森林）之间生产力的差异也很明显一天的时间。ECOSTRESS GPP还捕获了光合作用的昼夜循环中的季节性变化。这项研究证明了使用ECOSTRESS数据在一天中的不同时间生成瞬时GPP（即GPP用于ECOSTRESS数据的获取时间）的可行性。