Revealing the globally multiscale controls of environmental factors on carbon use efficiency.

The carbon use efficiency (CUE), which is the ratio of net primary production to gross primary production, is an essential element for detecting the terrestrial carbon cycle and ecosystem function. The spatial variation of CUE is controlled by environmental factors independently or interactively with different intensity. However, previous studies have mainly focused on the effect of climate on the local CUE at the sampling scale, while neglecting the effects of topography or soil on the global CUE, and even its spatially predictive model. In the study, the relative contributions of potentially influencing factors (i.e., climatic, topographic, and edaphic factors), and their interactions on the global CUE were analyzed using the combined methods of curvelet transform and geographical detector model, and the spatial model of CUE were established based on its relationships with influencing factors. The results showed that CUE values at the sampling scale were generally greater in the mid- and high-latitude regions than those in the low-latitude region, which was characterized by its spatial pattern at the large scale. Climate had the greater effects on CUE variations at the large scale, while topography was the main factor controlling CUE at the small or medium scale. However, the explanatory power of the interaction among factors on CUE was greater than any single factor, among which the interaction between climatic and topographical factors was the strongest at all scales. The CUE predication based on scale-dependent effects was more accurate than that based on the sampling scale especially in the high-latitude, and temperature and elevation was the main predictors. Based on the model, the spatial patterns of CUE under future scenarios with any climatic changes could be extracted. This study can further advance our understanding on spatial variation of CUE, and provide a unique insight for CUE prediction responding to climate changes.

Estimation of Winter Wheat Yield in Arid and Semiarid Regions Based on Assimilated Multi-Source Sentinel Data and the CERES-Wheat Model.

The farmland area in arid and semiarid regions accounts for about 40% of the total area of farmland in the world, and it continues to increase. It is critical for global food security to predict the crop yield in arid and semiarid regions. To improve the prediction of crop yields in arid and semiarid regions, we explored data assimilation-crop modeling strategies for estimating the yield of winter wheat under different water stress conditions across different growing areas. We incorporated leaf area index (LAI) and soil moisture derived from multi-source Sentinel data with the CERES-Wheat model using ensemble Kalman filter data assimilation. According to different water stress conditions, different data assimilation strategies were applied to estimate winter wheat yields in arid and semiarid areas. Sentinel data provided LAI and soil moisture data with higher frequency (<14 d) and higher precision, with root mean square errors (RMSE) of 0.9955 m2 m-2 and 0.0305 cm3 cm-3, respectively, for data assimilation-crop modeling. The temporal continuity of the CERES-Wheat model and the spatial continuity of the remote sensing images obtained from the Sentinel data were integrated using the assimilation method. The RMSE of LAI and soil water obtained by the assimilation method were lower than those simulated by the CERES-Wheat model, which were reduced by 0.4458 m2 m-2 and 0.0244 cm3 cm-3, respectively. Assimilation of LAI independently estimated yield with high precision and efficiency in irrigated areas for winter wheat, with RMSE and absolute relative error (ARE) of 427.57 kg ha-1 and 6.07%, respectively. However, in rain-fed areas of winter wheat under water stress, assimilating both LAI and soil moisture achieved the highest accuracy in estimating yield (RMSE = 424.75 kg ha-1, ARE = 9.55%) by modifying the growth and development of the canopy simultaneously and by promoting soil water balance. Sentinel data can provide high temporal and spatial resolution data for deriving LAI and soil moisture in the study area, thereby improving the estimation accuracy of the assimilation model at a regional scale. In the arid and semiarid region of the southeastern Loess Plateau, assimilation of LAI independently can obtain high-precision yield estimation of winter wheat in irrigated area, while it requires assimilating both LAI and soil moisture to achieve high-precision yield estimation in the rain-fed area.