Gap-filling Method Research Articles

Image similarity-based gap filling method can effectively enrich surface water mapping information

Satellite monitoring is an effective way to obtain the spatiotemporal information for surface water resources. However, due to problems such as clouds, cloud shadows, and sensor failures, there many gaps in satellite images, posing challenges for the accurate monitoring of surface water changes. Here, a new image-similarity-based gap-filling method using Landsat images, to obtain gapless surface water images, is proposed. The new method, based on a database composed of gapless images, uses the symmetrical difference method to determine the gapless image whose water pixels distribution is most similar to that of the gap image, finally replacing the gapped image pixels with gapless image pixels, to obtain a gap-filled water image. The performance of the new method is evaluated for arid and humid regions via numerical experiments and case verification. Then new method is compared with the existing gap-filling method, using the Aral Sea as an example. The new method’s average overall accuracy for surface water in arid areas is about 0.99, and that of surface water in humid areas is about 0.98. It predicted surface water more accurately than other methods, and with high accuracy (R2 = 0.92). This proposed gap-filling method provides an important technical solution to obtaining long-term high-quality observations of surface water. It can also provide a database of image for accurate water resource management.

A novel data gaps filling method for solar PV output forecasting

This study proposes a modified gaps filling method, expanding the column mean imputation method and evaluated using randomly generated missing values comprising 5%, 10%, 15%, and 20% of the original data on power output. The XGBoost algorithm was implemented as a forecasting model using the original and processed datasets and two sources of solar radiation data, namely, Shortwave Radiation (SWR) from Advanced Himawari Imager 8 (AHI-8) and Surface Solar Radiation Downward (SSRD) from ERA5 global reanalysis data. The accuracy of the two sets of forecasted power output was evaluated using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). Results show that by applying the proposed gap filling method and using SWR in forecasting solar photovoltaic (PV) output, the improvement in the RMSE and MAE values range from 12.52% to 24.30% and from 21.10% to 31.31%, respectively. Meanwhile, using SSRD, the improvement in the RMSE values range from 14.01% to 28.54% and MAE values from 22.39% to 35.53%. To further evaluate the accuracy of the proposed gap-filling method, the proposed method could be validated using different datasets and other forecasting methods. Future studies could also consider applying the said method to datasets with data gaps higher than 20%.

Variable combinatorial gap-filling method for single-cell RNA-seq data

Abstract With the increasing development of single-cell RNA sequencing technology, a huge amount of sequencing data has emerged. The use of computational methods to fill in the gene expression information in scRNA-seq data is not only an important guide for gene regulatory network construction, embryonic development, and neurological research in the brain but also provides an important basis for drug development and clinical medicine. In this paper, we propose a variable combination of single-cell gap-filling algorithms with high gap-filling accuracy and fast computation speed through the comprehensive study and analysis of image repair technology and single-cell gap-filling algorithm. The experiments demonstrate that the U-net-based gap-filling method proposed in this paper has high accuracy in recovering gene expression values, can reduce the analysis errors caused by dropout events, and applies to large-scale data sets. In summary, the variable combinatorial gap-filling method for single-cell RNA-seq data proposed in this paper can effectively improve the results of downstream analysis and promote the development of research in the field of RNA sequencing data.

CRYSTAL: A novel and effective method to remove clouds in daily nighttime light images by synergizing spatiotemporal information

Satellite nighttime light (NTL) images offer a valuable depiction of the rapidly changing world by revealing the presence of artificial illumination. Thus, daily NTL images are increasingly applied to monitor human dynamics and environmental events. However, data gaps caused by cloud contamination and low-quality observations inevitably impair the effectiveness of such applications. Although a temporal gap-filling method is employed in recent Black Marble NTL products to produce seamless images, the filled images are unsuitable for quantitative analysis. Therefore, we developed an effective method, named as Cloud Removing bY Synergizing spatio-TemporAL information (CRYSTAL), to generate cloud-free NTL images with satisfactorily accurate pixel brightness and spatial continuity. Simulation experiments show that CRYSTAL can produce more accurate results than the temporal gap-filling method in fifteen cities worldwide, with an average RMSE reduction of 33.69%. Images generated by CRYSTAL restore temporal variances in NTL and are thus suitable for multi-temporal quantitative analysis. CRYSTAL can reconstruct daily NTL time series by filling gaps using available partially clear images. Experiments in two cities demonstrated that the reconstructed time series had 31.85% more valid values than the original time series and effectively revealed urban dynamics during the early stages of the coronavirus disease 2019 pandemic. In summary, CRYSTAL is a novel and effective gap-filling method for the restoration of invalid NTL observations and has the potential to generate high-quality NTL data for use in future applications.

Using Enhanced Gap-Filling and Whittaker Smoothing to Reconstruct High Spatiotemporal Resolution NDVI Time Series Based on Landsat 8, Sentinel-2, and MODIS Imagery

Normalized difference vegetation index (NDVI) time series data, derived from optical images, play a crucial role for crop mapping and growth monitoring. Nevertheless, optical images frequently exhibit spatial and temporal discontinuities due to cloudy and rainy weather conditions. Existing algorithms for reconstructing NDVI time series using multi-source remote sensing data still face several challenges. In this study, we proposed a novel method, an enhanced gap-filling and Whittaker smoothing (EGF-WS), to reconstruct NDVI time series (EGF-NDVI) using Google Earth Engine. In EGF-WS, NDVI calculated from MODIS, Landsat-8, and Sentinel-2 satellites were combined to generate high-resolution and continuous NDVI time series data. The MODIS NDVI was employed as reference data to fill missing pixels in the Sentinel–Landsat NDVI (SL-NDVI) using the gap-filling method. Subsequently, the filled NDVI was smoothed using a Whittaker smoothing filter to reduce residual noise in the SL-NDVI time series. With reference to the all-round performance assessment (APA) metrics, the performance of EGF-WS was compared with the conventional gap-filling and Savitzky–Golay filter approach (GF-SG) in Fusui County of Guangxi Zhuang Autonomous Region. The experimental results have demonstrated that the EGF-WS can capture more accurate spatial details compared with GF-SG. Moreover, EGF-NDVI of Fusui County exhibited a low root mean square error (RMSE) and a high coefficient of determination (R2). In conclusion, EGF-WS holds significant promise in providing NDVI time series images with a spatial resolution of 10 m and a temporal resolution of 8 days, thereby benefiting crop mapping, land use change monitoring, and various ecosystems, among other applications.

Teasing out missing reactions in genome-scale metabolic networks through hypergraph learning

GEnome-scale Metabolic models (GEMs) are powerful tools to predict cellular metabolism and physiological states in living organisms. However, due to our imperfect knowledge of metabolic processes, even highly curated GEMs have knowledge gaps (e.g., missing reactions). Existing gap-filling methods typically require phenotypic data as input to tease out missing reactions. We still lack a computational method for rapid and accurate gap-filling of metabolic networks before experimental data is available. Here we present a deep learning-based method — CHEbyshev Spectral HyperlInk pREdictor (CHESHIRE) — to predict missing reactions in GEMs purely from metabolic network topology. We demonstrate that CHESHIRE outperforms other topology-based methods in predicting artificially removed reactions over 926 high- and intermediate-quality GEMs. Furthermore, CHESHIRE is able to improve the phenotypic predictions of 49 draft GEMs for fermentation products and amino acids secretions. Both types of validation suggest that CHESHIRE is a powerful tool for GEM curation to reveal unknown links between reactions and observed metabolic phenotypes.

Product, building, and infrastructure material stocks dataset for 337 Chinese cities between 1978 and 2020

Reliable city-level product, building, and infrastructure material stocks data are essential for understanding historical material use patterns, benchmarking material efficiency, and informing future recycling potentials. However, such urban material stocks data are often limited, due primarily to unavailable, inconsistent, or noncontinuous city-level statistics. Here, we provided such an Urban Product, Building, and Infrastructure Material Stocks (UPBIMS) dataset for China, a country that has undergone a remarkable urbanization process in the past decades, by collating different official statistics and applying various gap-filling methods. This dataset contains the stock of 24 materials contained in 10 types of products, buildings, and infrastructure in all 337 prefecture-level cities in China from 1978 to 2020. This quality controlled and unified dataset is the first of its kind with such a full coverage of all prefecture-level Chinese cities and can be used in a variety of applications, for example in urban geography, industrial ecology, circular economy, and climate change mitigation. Every piece of data is tagged with its source and the dataset will be periodically updated.

Reconstruction of Global Long-Term Gap-Free Daily Surface Soil Moisture from 2002 to 2020 Based on a Pixel-Wise Machine Learning Method

Global, long-term, gap-free, high quality soil moisture products are extremely important for hydrological monitoring and climate change research. However, soil moisture products produced from satellite observations have data gaps due to the limited capabilities of satellite orbit/swath and retrieval algorithms, which limit the regional and global applications of soil moisture data in hydrology and agriculture studies. To solve this problem, we proposed a gap-filling method to reconstruct a global gap-free surface soil moisture product by applying the machine learning (Random Forest) algorithm on a pixel-by-pixel basis, taking into account the nonlinear relationship between surface soil moisture and the related surface environmental variables. The gap-filling method was applied to the NN-SM surface soil moisture product, which has a fraction of data gaps of around 50% globally on a multi-year average. A global daily gap-free surface soil moisture dataset from 2002 to 2020 was then generated. The reconstructed values of several sub-regions after manually eliminating the original values were cross-verified with the original data, and this clearly demonstrated the reliability of the reconstruction method with the correlation coefficient (R) ranging between 0.770 and 0.918, the Root Mean Square Error (RMSE) between 0.057 and 0.082 m3/m3, the unbiased Root Mean Square Error (ubRMSE) between 0.053 and 0.081 m3/m3, and Bias between −0.012 and 0.008 m3/m3. The accuracy of the reconstructed surface soil moisture dataset was evaluated using in situ observations of surface soil moisture at 12 sites from the International Soil Moisture Network (ISMN) and the Long-Term Agroecosystem Research (LTAR) network, and the results showed good accuracy in terms of R (0.610), RMSE (0.067 m3/m3), ubRMSE (0.045 m3/m3) and Bias (0.031 m3/m3). Overall, the reconstructed surface soil moisture dataset retained the characteristics of the NN-SM product, such as high accuracy and good spatiotemporal pattern. However, with the advantage of continuous spatiotemporal coverage, it is more suitable for further applications in the analysis of global surface soil moisture trends, land surface hydrological processes, and land-atmosphere energy and water exchanges, etc.

Physics-Informed Deep Learning for Reconstruction of Spatial Missing Climate Information in the Antarctic

Understanding the influence of the Antarctic on the global climate is crucial for the prediction of global warming. However, due to very few observation sites, it is difficult to reconstruct the rational spatial pattern by filling in the missing values from the limited site observations. To tackle this challenge, regional spatial gap-filling methods, such as Kriging and inverse distance weighted (IDW), are regularly used in geoscience. Nevertheless, the reconstructing credibility of these methods is undesirable when the spatial structure has massive missing pieces. Inspired by image inpainting, we propose a novel deep learning method that demonstrates a good effect by embedding the physics-aware initialization of deep learning methods for rapid learning and capturing the spatial dependence for the high-fidelity imputation of missing areas. We create the benchmark dataset that artificially masks the Antarctic region with ratios of 30%, 50% and 70%. The reconstructing monthly mean surface temperature using the deep learning image inpainting method RFR (Recurrent Feature Reasoning) exhibits an average of 63% and 71% improvement of accuracy over Kriging and IDW under different missing rates. With regard to wind speed, there are still 36% and 50% improvements. In particular, the achieved improvement is even better for the larger missing ratio, such as under the 70% missing rate, where the accuracy of RFR is 68% and 74% higher than Kriging and IDW for temperature and also 38% and 46% higher for wind speed. In addition, the PI-RFR (Physics-Informed Recurrent Feature Reasoning) method we proposed is initialized using the spatial pattern data simulated by the numerical climate model instead of the unified average. Compared with RFR, PI-RFR has an average accuracy improvement of 10% for temperature and 9% for wind speed. When applied to reconstruct the spatial pattern based on the Antarctic site observations, where the missing rate is over 90%, the proposed method exhibits more spatial characteristics than Kriging and IDW.

Research on Side-Slope Monitoring by Integrating Terrestrial Laser Scanning and UAV-Based Photogrammetry

Abstract Side-slope deformation monitoring compares monitoring data from the same area over different periods and measures the deformation variables. Because of the gaps and coarseness of side-slope monitoring data, a side-slope monitoring method that integrates terrestrial laser scanning (TLS) and unmanned aerial vehicle (UAV)–based photogrammetry point clouds is proposed, aiming to solve the problem of slope monitoring in complex scenes. First, TLS and UAV-based photogrammetry point clouds are acquired. Then, the two types of point clouds are registered by an iterative closest point algorithm. Next, the data gap areas in the TLS point cloud are detected, and a gap-filling method is used to integrate the UAV-based photogrammetry point cloud with the TLS point cloud. Finally, side-slope deformation is detected based on a multiscale model-to-model cloud comparison algorithm. A side slope in Chenggong, Kunming, China, is taken as an example. The surface deformation of the side slope was monitored during January and June 2021. The experimental results show that the registration errors of the two-phase integration point cloud are 0.039 m and 0.035 m. The root mean square errors of the four ground checkpoints are 0.033 m and 0.038 m. Finally, the side slope is found to have deformed and formed a main deformation area, which shows that this side slope was in an active state.

A 21-year dataset (2000–2020) of gap-free global daily surface soil moisture at 1-km grid resolution

Global soil moisture estimates from current satellite missions are suffering from inherent discontinuous observations and coarse spatial resolution, which limit applications especially at the fine spatial scale. This study developed a dataset of global gap-free surface soil moisture (SSM) at daily 1-km resolution from 2000 to 2020. This is achieved based on the European Space Agency - Climate Change Initiative (ESA-CCI) SSM combined product at 0.25° resolution. Firstly, an operational gap-filling method was developed to fill the missing data in the ESA-CCI SSM product using SSM of the ERA5 reanalysis dataset. Random Forest algorithm was then adopted to disaggregate the coarse-resolution SSM to 1-km, with the help of International Soil Moisture Network in-situ observations and other optical remote sensing datasets. The generated 1-km SSM product had good accuracy, with a high correlation coefficent (0.89) and a low unbiased Root Mean Square Error (0.045 m3/m3) by cross-validation. To the best of our knowledge, this is currently the only long-term global gap-free 1-km soil moisture dataset by far.

A widely-used eddy covariance gap-filling method creates systematic bias in carbon balance estimates

Climate change mitigation requires, besides reductions in greenhouse gas emissions, actions to increase carbon sinks in terrestrial ecosystems. A key measurement method for quantifying such sinks and calibrating models is the eddy covariance technique, but it requires imputation, or gap-filling, of missing data for determination of annual carbon balances of ecosystems. Previous comparisons of gap-filling methods have concluded that commonly used methods, such as marginal distribution sampling (MDS), do not have a significant impact on the carbon balance estimate. By analyzing an extensive, global data set, we show that MDS causes significant carbon balance errors for northern (latitude >60^circ) sites. MDS systematically overestimates the carbon dioxide (CO_2) emissions of carbon sources and underestimates the CO_2 sequestration of carbon sinks. We also reveal reasons for these biases and show how a machine learning method called extreme gradient boosting or a modified implementation of MDS can be used to substantially reduce the northern site bias.

Improved cloudy-sky snow albedo estimates using passive microwave and VIIRS data

Land surface albedo (LSA) is an essential component of the surface radiation budget, and has been retrieved extensively as a basic remote sensing product; however, daily LSA products suffer from extensive data gaps primarily caused by cloud cover. Accordingly, several gap-filling methods were developed (e.g., spatiotemporal interpolation and data fusion with albedo climatology), although the traditional methods are limited by cloud scale and surface heterogeneity. Further, as the largest varying surface landscape feature, seasonal snow cover substantially influences LSA and represents a major uncertainty factor of gap recovery because previous studies failed to employ actual surface signals to capture such ephemeral but intense albedo changes under cloud cover. To address this issue, a three-step framework was proposed for estimating 1 km cloudy-sky LSA using passive microwave (PMW) data, albedo climatology, and Visible Infrared Imaging Radiometer Suite (VIIRS) clear-sky albedo: (1) All-sky snow albedo was estimated from PMW brightness temperatures using a statistical model, (2) Continuous albedo dynamics were generated by combining the all-sky snow albedo with snow-free climatological albedo, and (3) The 1 km cloudy-sky LSA was predicted after filtering 1 km VIIRS clear-sky LSA by the albedo dynamic series. PMW-derived snow albedo was assessed over the Contiguous US (CONUS), and the final 1 km cloudy-sky LSA was validated across 10 sites from SURFRAD and Core AmeriFlux in 2013. Based on the comparison with high-quality MODIS pixels, the estimated snow albedo yielded an overall RMSE of 0.064 over CONUS, with a bias of −0.010 (R2 = 0.845). The recovered 1 km cloudy-sky LSA produced RMSEs of 0.074 (0.137) for all (snow) samples, a significant improvement over the Global Land Surface Satellite (GLASS) gap-free albedo products especially on snow cases (p-value = 0.027). Corresponding RMSE in calculating surface net radiation was also decreased by 38.91 W·m−2; and anomalous snow samples were corrected as well. The temporal analysis and all-sky LSA mapping suggest that the recovered LSA has satisfactory spatiotemporal continuity, and successfully captured details of spatiotemporal variability, especially for ephemeral snow events. This study provides an innovative solution to recover gaps in LSA data, and considerably improves the LSA accuracy under cloud cover, which can inform snow melting modeling, hazard forecasting, and irrigation management.

High Spatial Resolution Gap-Free Global and Regional Ocean Color Products

Satellite ocean color products derived from the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (SNPP) and NOAA-20, and the Ocean and Land Colour Instrument (OLCI) on the Sentinel-3A (S3A) and Sentinel-3B (S3B) have been widely used for surveillance of the ocean environment and research on ocean physical, biological, biogeochemical, and ecological processes. However, either VIIRS or OLCI daily ocean color images are often incomplete in spatial coverage due to cloud cover, contamination of high sun glint, narrow swath width, high sensor-zenith angle, high solar-zenith angle, and/or other unfavorable retrieval conditions. Although merging daily ocean color images from multiple satellite sensors can help reduce the number of invalid pixels, gap-filling methods such as the Data Interpolating Empirical Orthogonal Function (DINEOF) are often used to reconstruct invalid pixels and generate gap-free images. The 9-km spatial resolution global gap-free ocean color data have been routinely produced by the NOAA Ocean Color Team and distributed through NOAA CoastWatch (https://coastwatch.noaa.gov/cw/index.html). In this study, we aim to develop and produce improved spatial resolution gap-free products, including chlorophyll-a (Chl-a) concentration, diffuse attenuation coefficient at the wavelength of 490 nm [ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">K_d</i> (490)], and suspended particulate matter (SPM) concentration for spatial resolutions of 0.5-km, 1-km, and 2-km. Two-sensor (VIIRS-SNPP and VIIRS-NOAA-20), three-sensor (two-sensor + OLCI-S3A), and four-sensor (three-sensor + OLCI-S3B) daily merged global Level-3 ocean color data are created and compared. It is found that by merging data from the two VIIRS sensors, ~38% more valid ocean product data are retrieved compared with a single sensor from either SNPP or NOAA-20. Adding OLCI-S3A to the two-sensor merged data can increase the number of valid pixels by ~12%; and adding OLCI-S3B to the three-sensor merged data can further increase the number of valid pixels by ~8%. The DINEOF method is applied to daily two-sensor, three-sensor, and four-sensor merged data to generate global 2-km resolution gap-free images. Results show that 2-km resolution gap-free images are able to resolve fine ocean features like coastal eddies and filaments, which are not available in the 9-km resolution images. While adding OLCI-S3A data significantly improves the three-sensor derived gap-free images over the two-sensor images, no significant enhancement is found in the four-sensor derived gap-free images by adding OLCI-S3B data. The DINEOF method is also applied to 1-km and 0.5-km resolution four-sensor merged Chl-a, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">K_d</i> (490), and SPM images in the Gulf of Mexico and U.S. west coast region. It is found that both 0.5-km and 1-km resolution images show more detailed ocean structures and features in coastal regions.

3D Hand Pose and Shape Estimation from Single RGB Image for Augmented Reality

In the realm of Human-Computer Interaction (HCI), the importance of hands cannot be overstated. Hands serve as a fundamental means of communication, expression, and interaction in the physical world. In recent years, Augmented Reality (AR) has emerged as a next-generation technology that seamlessly merges the digital and physical worlds, providing transformative experiences across various domains. In this context, accurate hand pose and shape estimation plays a crucial role in enabling natural and intuitive interactions within AR environments. Augmented Reality, with its ability to overlay digital information onto the real world, has the potential to revolutionize how we interact with technology. From gaming and education to healthcare and industrial training, AR has opened up new possibilities for enhancing user experiences. This study proposes an innovative approach for hand pose and shape estimation in AR applications. The methodology commences with the utilization of a pre-trained Single Shot Multi-Box (SSD) model for hand detection and cropping. The cropped hand image is then transformed into the HSV color model, followed by applying histogram equalization on the value band. To precisely isolate the hand, specific bounds are set for each band of the HSV color space, generating a mask. To refine the mask and diminish noise, contouring techniques are applied to the mask, and gap-filling methods are employed. The resultant refined mask is then combined with the original cropped image through logical AND operations to accurately delineate the hand boundaries. This meticulous approach ensures robust hand detection even in complex scenes. To extract pertinent features, the detected hand undergoes two concurrent processes. Firstly, the Scale-Invariant Feature Transform (SIFT) algorithm identifies distinctive keypoints on the hand's outer surface. Simultaneously, a pre-trained lightweight Convolutional Neural Network (CNN), namely MobileNet, is employed to extract 3D hand landmarks, the hand's center (middle finger metacarpophalangeal joint), and handedness information. These extracted features, encompassing hand keypoints, landmarks, center, and handedness, are aggregated and compiled into a CSV file for further analysis. A Gated Recurrent Unit (GRU) is then employed to process the features, capturing intricate dependencies between them. The GRU model successfully predicts the 3D hand pose, achieving high accuracy even in dynamic scenarios. The evaluation results for the proposed model are very promising that the Mean Per Joint Position Error in 3D (MPJPE) is 0.0596 between the predicted pose and the ground truth hand landmarks, while the Percentage of Correct Keypoints (PCK) is 95%. Upon predicting the hand pose, a mesh representation is employed to reconstruct the 3D shape of the hand. This mesh provides a tangible representation of the hand's structure and orientation, enhancing the realism and usability of the AR application. By integrating sophisticated detection, feature extraction, and predictive modeling techniques, this method contributes to creating more immersive and intuitive AR experiences, thereby fostering the seamless fusion of the digital and physical worlds.

A reduced latency regional gap-filling method for SMAP using random forest regression

The soil moisture active/passive (SMAP) mission represents a significant advance in measuring soil moisture from satellites. However, its large spatial-temporal data gaps limit the use of its values in near-real-time (NRT) applications. Considering this, the study uses NRT operational metadata (precipitation and skin temperature),together with some surface parameterization information, to feed into a random forest model to retrieve the missing values of the SMAP L3 soil moisture product. This practice was tested in filling the missing points for both SMAP descending (6:00 AM) and ascending orbits (6:00 PM) in a crop-dominated area from 2015 to 2019. The trained models with optimized hyper-parameters show the goodness of fit (R2 ≥ 0.86), and their resulting gap-filled estimates were compared against a range of competing products with in situ and triple collocation validation. This gap-filling scheme driven by low-latency data sources is first attempted to enhance NRT spatiotemporal support for SMAP L3 soil moisture.

Gap-filling of daily precipitation and streamflow time series: a method comparison at random and sequential gaps

ABSTRACT This study compared the performance of seven gap-filling methods in daily streamflow and precipitation data and assessed the maximum number of gap days on which the methods perform well. Random (occurring randomly in the data series) and sequential (consecutive days with missing record) gaps were considered. Results show that the type of gaps affects the performance of the methods for gap-filling streamflow and precipitation data. Concerning random gaps, the best methods for streamflow were autoregressive integrate moving average and spline interpolation. For precipitation, the best methods were inverse distance weighting and linear regression (LR). Regarding sequential gaps, LR and multiple regression perform well in gap-filling up to 60 consecutive days in streamflow series. The other methods perform well up to 15 consecutive gap days. For precipitation series, the methods performed well up to seven consecutive missing days. For longer gaps (15, 30, 45 and 60 days), the methods performed poorly.

A Single Image Deep Learning Approach to Restoration of Corrupted Landsat-7 Satellite Images

Remote sensing is increasingly recognized as a convenient tool with a wide variety of uses in agriculture. Landsat-7 has supplied multi-spectral imagery of the Earth's surface for more than 4 years and has become an important data source for a large number of research and policy-making initiatives. Unfortunately, a scan line corrector (SLC) on Landsat-7 broke down in May 2003, which caused the loss of up to 22 percent of any given scene. We present a single-image approach based on leveraging the abilities of the deep image prior method to fill in gaps using only the corrupt image. We test the ability of deep image prior to reconstruct remote sensing scenes with different levels of corruption in them. Additionally, we compare the performance of our approach with the performance of classical single-image gap-filling methods. We demonstrate a quantitative advantage of the proposed approach compared with classical gap-filling methods. The lowest-performing restoration made by the deep image prior approach reaches 0.812 in r2, while the best value for the classical approaches is 0.685. We also present the robustness of deep image prior in comparing the influence of the number of corrupted pixels on the restoration results. The usage of this approach could expand the possibilities for a wide variety of agricultural studies and applications.

Naive Bayes classification-based surface water gap-filling from partially contaminated optical remote sensing image

Optical remote sensing images are a common data sources for surface water monitoring, while they are easily contaminated by clouds, cloud shadows, terrain shadows, etc., resulting in spatial gaps in surface water images. This paper proposes a surface water gap-filling method based on Naive Bayes classification. It uses the historical cloud-free binary (water, non-water) surface water images as prior data and the uncontaminated pixels in the partially contaminated ternary (water, non-water, contaminated pixels) surface water image as evidence to identify the category of gap pixels to achieve the purpose of gap-filling. This method considers the relationship between disconnected water bodies and does not depend on terrain data. When the image is heavily covered by clouds, this method can also reconstruct the complete water extent accurately. Five study areas with different scenarios including rivers, lakes or reservoirs, are selected to evaluate the method. Results show that the average gap-filling accuracy in all five study areas is over 90 %. After gap-filling, the time series of surface water area presents a good correlation with the time series of water level (e.g., the coefficient of determinationR2 = 0.95 in the Dartmouth reservoir). The proposed method is proved effective in filling gaps caused by clouds, cloud shadows and terrain shadows in surface water image, and it would be suitable for high-frequency surface water monitoring and near real-time surface water mapping.

Detection of surface water temperature variations of Mongolian lakes benefiting from the spatially and temporally gap-filled MODIS data

• We applied an effective spatiotemporal gap-filled method to improve the LSWT from MODIS. • The long-term variations of the annual and seasonal mean LSWTs of Mongolian lakes were provided. • The LSWT of Mongolian lakes confirmed the phenomenon “global warming hiatus”. Lakes provide critical water resources for human activities and ecosystems, particularly in the Mongolian Plateau (MP), which is characterized by a dry climate and a harsh environment. As a region that is sensitive to anthropogenic warming, tracking lake surface water temperature (LSWT) changes in Mongolian lakes is crucial for understanding the consequences of a warming climate on lake ecosystems. However, the long-term monitoring of LSWT is restricted by the spatiotemporal gaps in the raw imagery of remote sensing-based land surface temperature (LST), e.g., the commonly used Moderate Resolution Imaging Spectroradiometer (MODIS) LST products. This study applied an improved gap-filling method by utilizing the discrete cosine transform-based penalized least squares (DCT-PLS) strategy in the spatial domain combined with the linear interpolation (LI) algorithm in the temporal domain. The method was applied to fill gaps in the LSWT imagery of 12 representative lakes across MP. The randomly sampled high-quality MODIS LSWT values in the spatial and temporal domains were excavated as false data gaps and considered “virtual true” validation datasets. The spatial validation results showed that the estimated LSWT for all the lake cases were comparable with the “virtual true” LSWT values, with the average values of the coefficient of determination, mean absolute error, mean square error, and root mean square error being 0.98, 0.38 °C, 0.45 °C, and 0.59 °C, respectively. Meanwhile, the error of nighttime LSWT results was relatively lower than that of daytime LSWT. For temporal interpolation validation, the LI algorithm exhibited relatively better performance and could more objectively indicate the variation in LSWT. Benefiting from the spatially and temporally well-constrained data, we analyzed the interannual and intra-annual change characteristics of the LSWTs of the 12 lakes. The long-term variations of annual and seasonal mean LSWTs in the 12 selected lakes exhibited no evident trends in 2000–2020, while presented apparent interannual fluctuations. The slight changes in the average LSWTs of the 12 selected lakes were in excellent synchronization with the surrounding LST derived from the reanalysis datasets, confirming the widely reported phenomenon of “global warming hiatus” that occurred in the early 21st century. This study improves the understanding of the LSWT variations in Mongolian lakes in response to global climate change. It has the potential to provide an effective approach for monitoring LSWT changes in other large-scale studies.