Recent increases in vegetation greenness over much of the world reflect increasing CO₂ globally and warming in cold areas. However, the strength of the response to both CO₂ and warming in those areas appears to be declining for unclear reasons, contributing to large uncertainties in predicting how vegetation will respond to future global changes. Here, we investigated the changes of satellite‐observed peak season absorbed photosynthetically active radiation (F_max) on the Tibetan Plateau between 1982 and 2016. Although climate trends are similar across the Plateau, we identified robust divergent responses (a greening of 0.31 ± 0.14% year⁻¹ in drier regions and a browning of 0.12 ± 0.08% year⁻¹ in wetter regions). Using an eco‐evolutionary optimality (EEO) concept of plant acclimation/adaptation, we propose a parsimonious modelling framework that quantitatively explains these changes in terms of water and energy limitations. Our model captured the variations in F_max with a correlation coefficient (r) of .76 and a root mean squared error of .12 and predicted the divergent trends of greening (0.32 ± 0.19% year⁻¹) and browning (0.07 ± 0.06% year⁻¹). We also predicted the observed reduced sensitivities of F_max to precipitation and temperature. The model allows us to explain these changes: Enhanced growing season cumulative radiation has opposite effects on water use and energy uptake. Increased precipitation has an overwhelmingly positive effect in drier regions, whereas warming reduces F_max in wetter regions by increasing the cost of building and maintaining leaf area. Rising CO₂ stimulates vegetation growth by enhancing water‐use efficiency, but its effect on photosynthesis saturates. The large decrease in the sensitivity of vegetation to climate reflects a shift from water to energy limitation. Our study demonstrates the potential of EEO approaches to reveal the mechanisms underlying recent trends in vegetation greenness and provides further insight into the response of alpine ecosystems to ongoing climate change.

Synthetic aperture radar (SAR) plays an essential role in earth observation and projection due to its capability to penetrate clouds, which makes it possible to monitor terrestrial surfaces under all weather conditions. Multiplicative noise often occurs in the SAR signal, hampering the retrieval of information from SAR imagery. Convolutional neural networks (CNNs) have been used in many computer vision tasks and are helpful in image denoising. However, current CNN-based denoising approaches inevitably lead to a “washed out” effect that loses spatial details. Another limitation is that most typical CNN-based denoising models require a noise-free image for training. To address these issues, we propose a novel end-to-end self-supervised SAR denoising model: Enhanced Noise2Noise (EN2N), which can be trained without a noise-free image. To enhance the quality of the result images, the perceptual features from a pre-learned CNN are introduced to restore the spatial details by a hybrid loss function. Experiments show that our proposed method outperforms the typical denoising methods in terms of noise reduction and feature preservation based on image quality metrics. Also, the new hybrid loss could enhance the spatial details significantly. The good performance maintains the robustness throughout time, which reduces the uncertainty in time-series SAR caused by random noise. Benefiting from optimization of graphics processing unit (GPU) and multi-threading, the proposed method has higher computation efficiency than traditional methods. This study demonstrates the great potential of using our self-supervised deep learning approaches for SAR image denoising in the future.

Net radiation plays an essential role in determining the thermal conditions of the Earth’s surface and is an important parameter for the study of land-surface processes and global climate change. In this paper, an improved satellite-based approach to estimate the daily net radiation is presented, in which sunshine duration were derived from the geostationary meteorological satellite (FY-2D) cloud classification product, the monthly empirical a_s and b_s Angstrom coefficients for net shortwave radiation were calibrated by spatial fitting of the ground data from 1997 to 2006, and the daily net longwave radiation was calibrated with ground data from 2007 to 2010 over the Heihe River Basin in China. The estimated daily net radiation values were validated against ground data for 12 months in 2008 at four stations with different underlying surface types. The average coefficient of determination (R²) was 0.8489, and the averaged Nash-Sutcliffe equation (NSE) was 0.8356. The close agreement between the estimated daily net radiation and observations indicates that the proposed method is promising, especially given the comparison between the spatial distribution and the interpolation of sunshine duration. Potential applications include climate research, energy balance studies and the estimation of global evapotranspiration.

Evapotranspiration (ET) links the water and carbon cycles in the atmosphere, hydrosphere, and biosphere. In this study, we develop an ET modelling framework based on the idea that the transpiration and carbon uptake are closely coupled, as predicted by the 'least-cost hypothesis' that canopy conductance acclimates to environmental variations. According to eco-evolutionary optimality theory, which has been previously applied in monitoring and modelling land-surface processes, the total costs (per unit carbon fixed) for maintaining transpiration and carboxylation capacities should be minimized. We calculate gross primary production (GPP) assuming that the light-and Rubisco-limited rates of photosynthesis, described by the classical biochemical model of photosynthesis, are coordinated on an approximately weekly time scale. Transpiration (T) is then calculated via acclimated canopy conductance, with no need for plant type-or biome-specific parameters. ET is finally calculated from T using an empirical function of light, temperature, soil water content and foliage cover to predict the T/ET ratio at each site. The GPP estimates were well supported by (weekly) GPP data at 20 widely distributed eddy-covariance flux sites (228 site-years), with correlation coefficients (r) = 0.81 and root-mean-square error (RMSE) = 18.7 gC week-1 and Nash-Sutcliffe efficiency (NSE) = 0.61. Predicted ET was also well supported, with r = 0.85, RMSE = 5.5 mm week-1 and NSE = 0.66. Estimated T/ET ratios (0.43-0.74) showed significant positive relationships to radiation, precipitation and green vegetation cover and negative relationships to temperature and modelled T (r = 0.84). Aspects of this framework could be improved, notably the estimation of T/ET. Nonetheless, we see the application of eco-evolutionary principles as a promising direction for water resources research, eliminating the uncertainty introduced by the need to specify multiple parameters, and leveraging the power of remotely sensed vegetation cover data as a key indicator of ecosystem function.

Land use/land cover (LULC) change has been recognized as one of the most important indicators to study ecological and environmental changes. Remote sensing provides an effective way to map and monitor LULC change in real time and for large areas. However, with the increasing spatial resolution of remote sensing imagery, traditional classification approaches cannot fully represent the spectral and spatial information from objects and thus have limitations in classification results, such as the “salt and pepper” effect. Nowadays, the deep semantic segmentation methods have shown great potential to solve this challenge. In this study, we developed an adaptive band attention (BA) deep learning model based on U-Net to classify the LULC in the Three Gorges Reservoir Area (TGRA) combining RapidEye imagery and topographic information. The BA module adaptively weighted input bands in convolution layers to address the different importance of the bands. By comparing the performance of our model with two typical traditional pixel-based methods including classification and regression tree (CART) and random forest (RF), we found a higher overall accuracy (OA) and a higher Intersection over Union (IoU) for all classification categories using our model. The OA and mean IoU of our model were 0.77 and 0.60, respectively, with the BA module and were 0.75 and 0.58, respectively, without the BA module. The OA and mean IoU of CART and RF were both below 0.51 and 0.30, respectively, although RF slightly outperformed CART. Our model also showed a reasonable classification accuracy in independent areas well outside the training area, which indicates the strong model generalizability in the spatial domain. This study demonstrates the novelty of our proposed model for large-scale LULC mapping using high-resolution remote sensing data, which well overcomes the limitations of traditional classification approaches and suggests the consideration of band weighting in convolution layers.