Recently, approaches based on deep learning are quite prevalent in the area of remote sensing scene classification. Though significant success has been achieved, these approaches are still subject to an excess of parameters and extremely dependent on a large quantity of labeled data. In this study, few-shot learning is used for remote sensing scene classification tasks. The goal of few-shot learning is to recognize unseen scene categories given extremely limited labeled samples. For this purpose, a novel task-adaptive embedding network is proposed to facilitate few-shot scene classification of remote sensing images, referred to as TAE-Net. A feature encoder is first trained on the base set to learn embedding features of input images in the pre-training phase. Then in the meta-training phase, a new task-adaptive attention module is designed to yield the task-specific attention, which can adaptively select informative embedding features among the whole task. In the end, in the meta-testing phase, the query image derived from the novel set is predicted by the meta-trained model with limited support images. Extensive experiments are carried out on three public remote sensing scene datasets: UC Merced, WHU-RS19, and NWPU-RESISC45. The experimental results illustrate that our proposed TAE-Net achieves new state-of-the-art performance for few-shot remote sensing scene classification.

Scene classification is a critical technology to solve the challenges of image search and image recognition. It has become an indispensable and challenging research topic in the field of remote sensing. At present, most scene classifications are solved by deep neural networks. However, existing methods require large-scale training samples and are not suitable for actual scenarios with only a few samples. For this reason, a framework based on metric learning and local descriptors (MLLD) is proposed to enhance the classification effect of remote sensing scenes on the basis of few-shot. Specifically, MLLD adopts task-level training that is carried out through meta-learning, and meta-knowledge is learned to improve the model’s ability to recognize different categories. Moreover, Manifold Mixup is introduced by MLLD as a feature processor for the hidden layer of deep neural networks to increase the low confidence space for smoother decision boundaries and simpler hidden layer representations. In the end, a learnable metric is introduced; the nearest category of the image is matched by measuring the similarity of local descriptors. Experiments are conducted on three public datasets: UC Merced, WHU-RS19, and NWPU-RESISC45. Experimental results show that the proposed scene classification method can achieve the most advanced results on limited datasets.

GaoFen6 (GF-6), successfully launched on June 2, 2018, is the sixth satellite of the High-Definition Earth observation system (HDEOS). Although GF-6 is the first high-resolution satellite in China to achieve precise agricultural observation, it will be widely used in many other domains, such as land survey, natural resources management, emergency management, ecological environment and so on. The GF-6 was not equipped with an onboard calibration instrument, so on-orbit radiometric calibration is essential. This paper aimed at the on-orbit radiometric calibration of the wide field of view camera (WFV) onboard GF-6 (GF-6/WFV) in multispectral bands. Firstly, the radiometric capability of GF-6/WFV is evaluated compared with the Operational Land Imager (OLI) onboard Landsat-8, Multi Spectral Instrument (MSI) onboard Sentinel-2 and Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Terra, which shows that GF-6/WFV has an obvious attenuation. Consequently, instead of vicarious calibration once a year, more frequent calibration is required to guarantee its radiometric consistency. The cross-calibration method based on the Badain Jaran Desert site using the bi-directional reflectance distribution function (BRDF) model calculated by Landsat-8/OLI and ZY-3/Three-Line Camera (TLC) data is subsequently applied to GF-6/WFV and much higher frequencies of calibration are achieved. Finally, the cross-calibration results are validated using the synchronized ground measurements at Dunhuang test site and the uncertainty of the proposed method is analyzed. The validation shows that the relative difference of cross-calibration is less than 5% and it is satisfied with the requirements of cross-calibration.

The wide field of view (WFV) is an optical imaging sensor on-board the Gao Fen 1 (GF-1). The WFV lacks an on-board calibrator, so on-orbit radiometric calibration is required. Zhong et al. proposed a method for cross-calibrating the charge-coupled device on-board the Chinese Huan Jing 1 (HJ-1/CCD) that can be applied to the GF-1/WFV. However, the accuracy is limited because of the wider radiometric dynamic range and the higher spatial resolution of the GF-1/WFV. Therefore, Landsat-8 Operational Land Imager (OLI) imagery with a radiometric resolution similar to that of the GF-1/WFV and DEM extracted from ZY-3 three-line array panchromatic camera (TLC) with a higher spatial resolution were used. A calibration site with uniform surface material and a natural topographic variation was selected, and a model of this site’s bidirectional reflectance distribution function (BRDF) was developed. The model has excellent agreement with the real situation, as shown by the comparison of the simulations to the actual OLI surface reflectance. Then, the model was used to calibrate the WFV. Compared with the TOA reflectance from synchronized Landsat-8/OLI images, all errors calculated with the calibration coefficients retrieved in this paper are less than 5%, much less than the errors calculated with the calibration coefficients given by the China Centre for Resource Satellite Data and Application (CRESDA).

Vegetation plays a fundamental role within terrestrial ecosystems, serving as a cornerstone of their functionality. Presently, these crucial ecosystems face a myriad of threats, including deforestation, overgrazing, wildfires, and the impact of climate change. The implementation of remote sensing for monitoring the status and dynamics of vegetation ecosystems has emerged as an indispensable tool for advancing ecological research and effective resource management. This study takes a comprehensive approach by integrating ecosystem monitoring indicators and aligning them with the objectives of SDG15. We conducted a thorough analysis by leveraging global 500 m resolution products for vegetation Leaf Area Index (LAI) and land cover classification spanning the period from 2016 to 2020. This encompassed the calculation of annual average LAI, identification of anomalies, and evaluation of change rates, thereby enabling a comprehensive assessment of the global status and transformations occurring within major vegetation ecosystems. In 2020, a discernible rise in the annual Average LAI of major vegetation ecosystems on a global scale became evident when compared to data from 2016. Notably, the ecosystems demonstrating a slight increase in area constituted the largest proportion (34.23%), while those exhibiting a significant decrease were the least prevalent (6.09%). Within various regions, such as Eastern Europe, Central Africa, and South Asia, substantial increases in both forest ecosystem area and annual Average LAI were observed. Furthermore, Eastern Europe and Central America recorded significant expansions in both grassland ecosystem area and annual average LAI. Similarly, regions experiencing notable growth in both cropland ecosystem areas and annual average LAI encompassed Southern Africa, Northern Europe, and Eastern Africa.

Land surface temperature (LST) is an important input parameter to characterize urban environmental heat change. Existing satellite-borne thermal infrared sensor technology cannot completely support the applications using high spatial resolution LST, such as analysis of urban thermal environment and energy consumption assessment. Downscaling LST is an alternative method to retrieve LST of high spatial resolution. In this paper, we propose an improved multi-factor geographically weighted regression (MFGWR) algorithm for LST downscaling. More factors were incorporated into geographically weighted regression method by taking into account different land covers and temporal variation so that the downscaled LST at urban areas with complicated land cover at various seasons was improved. It was applied to four urban areas with large difference on land cover at different seasons. Taking into account different factors, the temperature distribution of MFGWR reproduced additional spatial detail. Compared with the major statistical LST downscaling methods including thermal image sharpening algorithm (TsHarp), multiple scale factors with adaptive thresholds algorithm (MSFAT), support vector machine regression combined with gradient boosting (SVR-GB), and GWR, MFWGR showed a stable performance and higher accuracy.