Deriving gapless CO2 concentrations using a geographically weighted neural network: China, 2014–2020

Seasonal and Diurnal Variations in XCO2 Characteristics in China as Observed by OCO‐2/3 Satellites: Effects of Land Cover and Local Meteorology

“…Our investigation that using direct satellite-based observations to study the XCO 2 variation in seasonal and diurnal cycles amplitude among different biomes gives some new insight into carbon cycle amplitude, to some extent, could also reduce the uncertainties caused by the new assumptions of source/sink inversion work (Kou et al, 2023) or machine learning methods (Sheng et al, 2023;Y. Wang, Yuan, et al, 2023;L. Zhang et al, 2022b).…”

Improved Consistency of Satellite XCO2 Retrievals Based on Machine Learning

“…High‐quality observations requiring strict cloud filtering only account for a small fraction of total observations (Taylor et al., 2022) and have seasonal variability, leading to issues of spatial representation and uncertainty in flux inversion estimates (Houweling et al., 2015). Recent studies have utilized machine learning (ML) models to derive global full‐coverage XCO 2 maps based on data from GOSAT, OCO‐2, or both (He et al., 2022; Siabi et al., 2019; Wang et al., 2023; Zhang & Liu, 2023; Zhang et al., 2022), facilitating analysis of carbon dynamics (He et al., 2022). However, these seamless data sets lack sufficient accuracy and do not provide uncertainty estimates used for observational inputs in carbon flux inversion.…”

Seamless mapping of long-term (2010–2020) daily global XCO2 and XCH4 from GOSAT, OCO-2, and CAMS-EGG4 with a spatiotemporally self-supervised fusion method

“…A lot of efforts have been made to generate seamless XCO2 and XCH4 products for GOSAT and OCO-2. Initially, interpolationbased methods are widely utilized, such as the fixed rank kriging interpolation (Katzfuss and Cressie, 2011), semantic kriging interpolation (Bhattacharjee et al, 2014), and space-time kriging interpolation (He et al, 2020;Li et al, 2022). However, the interpolated results are usually performed at coarse spatial resolutions (e.g., 1°) and tend to show high uncertainties and oversmoothed distribution due to the extreme sparsity of original data.…”

“…However, the interpolated results are usually performed at coarse spatial resolutions (e.g., 1°) and tend to show high uncertainties and oversmoothed distribution due to the extreme sparsity of original data. At present, data fusion techniques (He et al, 2022a, b;Zhang et al, 2022;Zhang and Liu, 2023;Siabi et al, 2019) have emerged as new methods to acquire full-coverage products for GOSAT and OCO-2 at a high spatial resolution, which absorb advantages from multisource data. Generally, these methods exploited machine learning algorithms to train an end-to-end fusion function with multiple seamless data (e.g., model and reanalysis) as inputs.…”

“…Generally, these methods exploited machine learning algorithms to train an end-to-end fusion function with multiple seamless data (e.g., model and reanalysis) as inputs. For example, Siabi et al (2019) employed multi-layer perceptron and eight environmental variables (e.g., net primary productivity and leaf area index) to map full-coverage XCO2 in Iran; He et al (2022b) established seamless results over China using the OCO-2 XCO2 product, CarbonTracker model data, and auxiliary co-variates based on the light gradient boosting machine; Zhang et al (2022) proposed a geographically weighted neural network to produce full-coverage XCO2 product across China by fusing the datasets from OCO-2, CAMS-EGG4 (reanalysis), and ERA5; and Zhang and Liu (2023) adopted multiple datasets, e.g., EnviSat, GOSAT, OCO-2, CarbonTracker, and ERA5, and obtained long-term seamless XCO2 product in China through a finely devised neural network. These data fusion approaches provided high-quality results with seamless distribution and greatly enhance the data availability for GOSAT and OCO-2.…”