Per-pixel Basis Research Articles

A Comprehensive Method for Example-Based Color Transfer with Holistic–Local Balancing and Unit-Wise Riemannian Information Gradient Acceleration

Color transfer, an essential technique in image editing, has recently received significant attention. However, achieving a balance between holistic color style transfer and local detail refinement remains a challenging task. This paper proposes an innovative color transfer method, named BHL, which stands for Balanced consideration of both Holistic transformation and Local refinement. The BHL method employs a statistical framework to address the challenge of achieving a balance between holistic color transfer and the preservation of fine details during the color transfer process. Holistic color transformation is achieved using optimal transport theory within the generalized Gaussian modeling framework. The local refinement module adjusts color and texture details on a per-pixel basis using a Gaussian Mixture Model (GMM). To address the high computational complexity inherent in complex statistical modeling, a parameter estimation method called the unit-wise Riemannian information gradient (uRIG) method is introduced. The uRIG method significantly reduces the computational burden through the second-order acceleration effect of the Fisher information metric. Comprehensive experiments demonstrate that the BHL method outperforms state-of-the-art techniques in both visual quality and objective evaluation criteria, even under stringent time constraints. Remarkably, the BHL method processes high-resolution images in an average of 4.874 s, achieving the fastest processing time compared to the baselines. The BHL method represents a significant advancement in the field of color transfer, offering a balanced approach that combines holistic transformation and local refinement while maintaining efficiency and high visual quality.

Pixel-wise matching cost function for robust light field depth estimation

Depth estimation has significant potential in industrial applications, such as robot navigation. Light Field (LF) technology captures both spatial and angular information of a scene, enabling precise depth acquisition. Stereo matching technology has been commonly used for this task in LF imaging. However, previous methods typically compute the matching cost by simply shifting images with predefined offsets, causing poor performance in regions with sudden depth distribution changes (e.g., edges). To address this issue, we propose a Pixel-wise Matching Cost Function (PMCF) to estimate depth in units of pixel, then design a pixel-based network, termed PixelNet, for depth prediction, which achieves top rankings on the HCI 4D light field benchmark. Specifically, our cost function consists of two modules: the range search strategy and the modulation mechanism. The range search strategy enables the network to iteratively optimize depth on a per-pixel basis, while the modulation mechanism effectively handles scene noise and occlusions. By integrating these two modules, our function enables the generation of depth maps with sharper edges and smoother surfaces, compared with previous methods. Finally, we design a pixel-based network and apply it to both synthetic and real-world scenes, demonstrating that our method outperforms state-of-the-art methods in both scenes. Furthermore, PixelNet can also function as a post-processing network that can be seamlessly integrated with existing methods for depth refinement.

A new procedure to find the optimum deconvolution kernel to deblur satellite images

ABSTRACT The digital number of a given pixel in a satellite image does not only measure the radiance originating from the area of the Earth’s surface represented by this pixel, but also measures radiance originating from surrounding areas represented by adjacent pixels. This adjacency effect, also known as image blurring, is a fundamental limitation of satellite images, and introduces errors in all image processing techniques that quantify the properties of the Earth’s surface on a per-pixel basis. Image deblurring via deconvolution, is a traditional technique to reduce the adjacency effect, which has proven effective in a variety of remote sensing applications. For example, it has provided an almost fourfold reduction of the adjacency error, when using support vector machine classifiers to estimate land cover proportion at a subpixel level. The remote sensing literature usually assumes that satellite images should be deconvolved with a kernel defined by the sensor’s Point Spread Function PSF(x, y), overlooking the results of three seminal empirical studies which systematically demonstrated that the optimum deconvolution kernel is a shrunk PSF of the form PSF(x/β, y/β), where β < 1 is the optimum shrink factor. The current empirical procedure to find the optimum shrink factor is laborious and restrictive since it requires suitable satellite images of much higher resolution, such that their Ground Sampling Distance (GSD) is equal to the GSD of the sensor of interest divided by an integer number much greater than one. A new theoretical procedure that uses synthetic edge images to find the optimum shrink factor is proposed and applied to the same cases considered by the current empirical procedure, showing that the same results are obtained. The new procedure is much simpler to apply and more accurate and versatile, opening a variety of research paths on satellite image deblurring.

Comparison of Three Remote Sensing Indices in Revealing the Vegetation Growth Dynamics in Nepal from 2000 to 2020

Remote sensing indices have been widely used to monitor the vegetation growth dynamics induced by climate change and human activities, and yet the consistency of the vegetation dynamics revealed by different remote sensing indices in mountains is unclear. Using Nepal as a case study, this study explored the spatial-termporal consistencies of the three widely-used remote sensing indices （i.e., normalized difference vegetation index （NDVI）, leaf area index （LAI）, and net primary production （NPP）） in quantifying the vegetation growth dynamics in mountainous regions. The results indicated that the spatial distributions of the multi-year mean estimates varied greatly by remote sensing index, especially in the low-altitude regions. The maximum NDVI, LAI, and NPP occurred in the low, medium, and high mountain regions, respectively. Although all three indices showed an overall increasing tendency from a long-term perspective, the area percentage of the lands with a significant trend was obviously larger in NDVI （82%） than that in NPP （58%） and LAI （56%）. In addition, the land area percentages with vegetation growth enhancement decreased gradually by the rise of altitude for both the NDVI and LAI indices but decreased after an increase for the NPP index. Only 9.6% of the lands showed consistent long-term trends （with the same change directions and significant levels） in the three indices on a per-pixel basis. Our findings highlight the large uncertainties of remote sensing indices in monitoring vegetation growth dynamics in mountainous areas, and the importance of developing reinforced remote sensing products in future efforts.

Deformation Characteristics and Activation Dynamics of the Xiaomojiu Landslide in the Upper Jinsha River Basin Revealed by Multi-Track InSAR Analysis

The upper Jinsha River, located in a high-mountain gorge with complex geological features, is highly prone to large-scale landslides, which could result in the formation of dammed lakes. Analyzing the movement characteristics of the typical Xiaomojiu landslide in this area contributes to a better understanding of the dynamics of landslides in the region, which is of great significance for landslide risk prediction and analysis. True displacement data on the surface of landslides are crucial for understanding the morphological changes in landslides, providing fundamental parameters for dynamic analysis and risk assessment. This study proposes a method for calculating the actual deformation of landslide bodies based on multi-track Interferometric Synthetic Aperture Radar (InSAR) deformation data. It iteratively solves for the optimal true deformation vector of the landslide on a per-pixel basis under a least-squares constraint based on the assumption of consistent displacement direction among adjacent points on the landslide surface. Using multi-track Sentinel data from 2017 to 2023, the line of sight (LOS) accumulative de-formation of the Xiaomojiu landslide was obtained, with a maximum LOS deformation of −126 mm/year. The true surface displacement of the Xiaomojiu landslide after activation was calculated using LOS deformation. The development of two rotational sub-slipping zones on the landslide body is inferred based on the distribution of actual displacements along the central profile line. Analysis of temporal changes in water body area data revealed that the Xiaomojiu landslide was activated after a barrier lake event and continuously moved due to the influence of higher water levels’ in the river channel. In conclusion, the proposed method can be applied to calculate the true surface displacement of landslides with complex mechanisms for analyzing the movement status of landslide bodies. Furthermore, the spatiotemporal analysis of the Xiaomojiu landslide characteristics can support analyzing the mechanisms of similar landslides in the Jinsha River Basin.

Transforming a Non-Differentiable Rasterizer into a Differentiable One with Stochastic Gradient Estimation

We show how to transform a non-differentiable rasterizer into a differentiable one with minimal engineering efforts and no external dependencies (no Pytorch/Tensorflow). We rely on Stochastic Gradient Estimation, a technique that consists of rasterizing after randomly perturbing the scene's parameters such that their gradient can be stochastically estimated and descended. This method is simple and robust but does not scale in dimensionality (number of scene parameters). Our insight is that the number of parameters contributing to a given rasterized pixel is bounded. Estimating and averaging gradients on a per-pixel basis hence bounds the dimensionality of the underlying optimization problem and makes the method scalable. Furthermore, it is simple to track per-pixel contributing parameters by rasterizing ID- and UV-buffers, which are trivial additions to a rasterization engine if not already available. With these minor modifications, we obtain an in-engine optimizer for 3D assets with millions of geometry and texture parameters.

Spatial characterization of drought through CHIRPS and a station-based dataset in the Eastern Mediterranean

Abstract. Drought is a natural hazard which occurs in all climatic zones and affects different sectors, such as irrigation, energy, water supply, and ecology. Monitoring and predicting drought are pressing challenges, as drought is becoming more common and severe owing to the impacts of climate change and increased climatic variability. However, in many areas of the globe, the temporal and spatial characterization of droughts and drought severities are hindered by a lack of reliable, locally-measured long-term data and unevenly distributed, erratic meteorological stations. In this situation, remote sensing datasets such as Climate Hazards Group InfraRed Precipitation with Station Data (CHIRPS) can offer valuable insights into long-term developments and the spatial characteristics of droughts. Nonetheless, it is often uncertain to which extent data such as CHIRPS succeeds in representing local dynamics and how this varies between geographical regions and climate zones. In this analysis, we aim to evaluate spatial drought conditions over the Seyhan River basin in Turkey in the Eastern Mediterranean. Using the Standardized Precipitation Index (SPI) as a drought index, the applicability of CHIRPS as a long-term satellite precipitation product for drought monitoring is investigated. We compare two spatial representations of the SPI: one derived on a per-pixel basis from CHIRPS data since 1981, the other based on data from 19 meteorological stations scattered across the basin, which was spatialized using inverse distance weighted interpolation (IDW). Our results offer insights into the relative accuracy of CHIRPS data and avenues towards optimizing the quality of spatial drought characterization.

Event encryption: rethinking privacy exposure for neuromorphic imaging

Bio-inspired neuromorphic cameras sense illumination changes on a per-pixel basis and generate spatiotemporal streaming events within microseconds in response, offering visual information with high temporal resolution over a high dynamic range. Such devices often serve in surveillance systems due to their applicability and robustness in environments with high dynamics and harsh lighting, where they can still supply clearer recordings than traditional imaging. In other words, when it comes to privacy-relevant cases, neuromorphic cameras also expose more sensitive data and pose serious security threats. Therefore, asynchronous event streams necessitate careful encryption before transmission and usage. This work discusses several potential attack scenarios and approaches event encryption from the perspective of neuromorphic noise removal, in which we inversely introduce well-crafted noise into raw events until they are obfuscated. Our evaluations show that the encrypted events can effectively protect information from attacks of low-level visual reconstruction and high-level neuromorphic reasoning, and thus feature dependable privacy-preserving competence. The proposed solution gives impetus to the security of event data and paves the way to a highly encrypted technique for privacy-protective neuromorphic imaging.

OHO: A Multi-Modal, Multi-Purpose Dataset for Human-Robot Object Hand-Over.

In the context of collaborative robotics, handing over hand-held objects to a robot is a safety-critical task. Therefore, a robust distinction between human hands and presented objects in image data is essential to avoid contact with robotic grippers. To be able to develop machine learning methods for solving this problem, we created the OHO (Object Hand-Over) dataset of tools and other everyday objects being held by human hands. Our dataset consists of color, depth, and thermal images with the addition of pose and shape information about the objects in a real-world scenario. Although the focus of this paper is on instance segmentation, our dataset also enables training for different tasks such as 3D pose estimation or shape estimation of objects. For the instance segmentation task, we present a pipeline for automated label generation in point clouds, as well as image data. Through baseline experiments, we show that these labels are suitable for training an instance segmentation to distinguish hands from objects on a per-pixel basis. Moreover, we present qualitative results for applying our trained model in a real-world application.

First Nighttime Light Spectra by Satellite—By EnMAP

For the first time, nighttime VIS/NIR—SWIR (visible and near-infrared—shortwave infrared) spectra from a satellite mission have been analyzed using the EnMAP (Environmental Mapping and Analysis Program) high-resolution imaging spectrometer. This article focuses on the spectral characteristics. Firstly, we checked the spectral calibration of EnMAP using sodium light emissions. Here, By applying a newly devised general method, we estimated shifts of +0.3nm for VIS/NIR and −0.2nm for SWIR; the uncertainties were found to be within the range of [−0.4nm,+0.2nm] for VIS/NIR and [−1.2nm,+1.0nm] for SWIR. These results emphasize the high accuracy of the spectral calibration of EnMAP and illustrate the feasibility of methods based on nighttime Earth observations for the spectral calibration of future nighttime satellite missions. Secondly, by employing a straightforward general method, we identified the dominant lighting types and thermal emissions in Las Vegas, Nevada, USA, on a per-pixel basis, and we considered the consistency of the outcomes. The identification and mapping of different types of LED (light-emitting diode) illuminations were achieved—with 75% of the identified dominant lighting types identified in VIS/NIR—as well as high- and low-pressure sodium and metal halide, which made up 22% of the identified dominant lighting types in VIS/NIR and 29% in SWIR and other illumination sources, as well as high temperatures, where 33% of the identified dominant emission types in SWIR were achieved from space using EnMAP due to the elevated illumination levels in the observed location. These results illustrate the feasibility of the precise identification of lighting types and thermal emissions based on nighttime high-resolution imaging spectroscopy satellite products; moreover, they support the specification of spectral characteristics for upcoming nighttime missions.

High-frame rate homography and visual odometry by tracking binary features from the focal plane

Robotics faces a long-standing obstacle in which the speed of the vision system’s scene understanding is insufficient, impeding the robot’s ability to perform agile tasks. Consequently, robots must often rely on interpolation and extrapolation of the vision data to accomplish tasks in a timely and effective manner. One of the primary reasons for these delays is the analog-to-digital conversion that occurs on a per-pixel basis across the image sensor, along with the transfer of pixel-intensity information to the host device. This results in significant delays and power consumption in modern visual processing pipelines. The SCAMP-5—a general-purpose Focal-plane Sensor-processor array (FPSP)—used in this research performs computations in the analog domain prior to analog-to-digital conversion. By extracting features from the image on the focal plane, the amount of data that needs to be digitised and transferred is reduced. This allows for a high frame rate and low energy consumption for the SCAMP-5. The focus of our work is on localising the camera within the scene, which is crucial for scene understanding and for any downstream robotics tasks. We present a localisation system that utilise the FPSP in two parts. First, a 6-DoF odometry system is introduced, which efficiently estimates its position against a known marker at over 400 FPS. Second, our work is extended to implement BIT-VO—6-DoF visual odometry system which operates under an unknown natural environment at 300 FPS.

Cropland Mapping in Tropical Smallholder Systems with Seasonally Stratified Sentinel-1 and Sentinel-2 Spectral and Textural Features

Mapping arable field areas is crucial for assessing agricultural productivity but poses challenges in sub-Saharan agroecosystems because of diverse crop calendars, small and irregularly shaped fields, persistent cloud cover, and lack of high-quality model training data. This study proposes several methodological improvements to overcome these challenges. Specifically, it utilizes long-term MODIS data to stratify finer Sentinel-2 reflectance and Sentinel-1 backscatter image features on a per-pixel basis. It also incorporates texture features and employs a machine learning approach with over 300,000 samples. The eastern region of Ghana was stratified into seven seasonal strata exhibiting distinct vegetation seasonality, capturing diversity in crop calendars, using long-term MODIS (2001–2009) normalized difference vegetation index phenology. Three years (2017–2019) of Sentinel-1 and Sentinel-2 original bands at 20 m were composited into dry and wet seasonal features according to the strata, from which spectral, polarimetric, and texture features were extracted. The field boundaries were digitized using PlanetScope images (2018–2019). Random Forest classifier with 10-fold cross-validation and recursive feature elimination was used for feature selection and model building. Including topographic variables, out of 137 image features, only 11 features were found important. Sentinel-2 SWIR-based spectral features were most important, followed by Sentinel-1 polarimetric (VV) and elevation features. Half of the 11 features were variance texture features, followed by spectral features. The Random Forest classifier produced a 0.78 AUC score with overall precision, recall, and F1-score of 0.96, 0.78, and 0.85, respectively. While the precision for both classes was &gt;0.90, the recall rate for arable areas was half that of non-arable areas. Future studies could improve the technical workflow with reliable balanced sampling, narrowband hyperspectral images, and fully polarized SAR images.

Design of a line-of-sight fluorescence-based imaging diagnostic for classification of microbe species

Fluorescence imaging of certain biochemicals, including flavins and pyridine nucleotides, has utility in characterizing the metabolic state of tissue and in discriminating between microbial species. There is significant clinical utility in this class of imaging techniques but most measurements reported to date require specialized training and equipment rendering most implementations unsuitable for routine medical imaging. Here, a low-cost and robust imaging technique is designed using ultraviolet-induced fluorescence of pyridine nucleotides (primarily NADH) and flavins (primarily FAD) in microbial samples. The diagnostic is optimized to distinguish between different microbial species based on previously reported spectral data using a ratiometric imaging approach. A detailed performance analysis is provided that relates the measured fluorescence intensity ratio (FIR) to the relative concentration ratio of NADH to FAD using a simplified spectroscopic model. Analysis suggests the technique is sensitive to changes in the NADH/FAD concentration ratio over several orders of magnitude, with better than 10% FIR precision on a per-pixel basis for microbial smears as thin as 10 s of microns at a resolution of 30 mm−1 and exposures of 20 ms. Representative microbe samples from eight species were imaged to demonstrate the proposed technique. Results show that the FIR varies by an order of magnitude across different species but the intra-species variation is only ∼5% for the conditions used here. An additional imaging band may be necessary to classify species that contain red pigments or bacteriochlorophyll. Radiative trapping was discussed as a possible limitation of the technique, but no clear evidence for radiative trapping was observed here. Overall, the results suggest that the proposed approach is feasible for rapid, low-cost, and robust characterization of microbial samples.

Mapping tea plantation area using phenology algorithm, time-series Sentinel-2 and Landsat images

ABSTRACT Tea plants are evergreen broad-leaved perennial shrubs, and their spectral characteristics are very similar to those of other evergreen vegetation, making it difficult to distinguish them. Currently, the most commonly used method of tea plantation extraction is machine learning, which classifies tea plantation through various feature combinations and algorithms. The disadvantage of these methods is that they require a large number of local training samples, making it challenging to produce an accurate model applicable to a large region. Furthermore, complex feature combinations and indicators may result in over-fitting, reducing the accuracy of the results. Therefore, it is necessary to develop a new algorithm suitable for tea plantation extraction in extensive regions. This paper uses Shihe District, Henan Province, China as a case in 2019, combined with Landsat-7/8 and Sentinel-2A/B images, and develops a new phenological-based algorithm to extract the tea plantation area. Firstly, we generated an evergreen vegetation map. Secondly, based on high-quality time series curves, the tea plant growth period was divided into seven parts to extract phenological indicators for classification. Finally, the tea plantation in the study area was extracted on a per-pixel basis. The overall accuracy of the algorithm is 87.59%, and the Kappa coefficient is 0.80. This study demonstrates the potential of the phenology-algorithms in extracting tea plantation areas and provides an advanced scheme and scientific basis for extracting tea plantation in other years, and also offers a reference for identifying tea plantation in other regions. Additionally, this paper generated a map of classified phenological indicators to provide guidance for monitoring tea plants growth.

Cloud Climatologies from Global Climate Models—A Comparison of CMIP5 and CMIP6 Models with Satellite Data

Abstract Simulating clouds with global climate models is challenging as the relevant physics involves many nonlinear processes covering a wide range of spatial and temporal scales. As key components of the hydrological cycle and the climate system, an evaluation of clouds from models used for climate projections is an important prerequisite for assessing the confidence in the results from these models. Here, we compare output from models contributing to phase 6 of the Coupled Model Intercomparison Project (CMIP6) with satellite data and with results from their predecessors (CMIP5). We use multiproduct reference datasets to estimate the observational uncertainties associated with different sensors and with internal variability on a per-pixel basis. Selected cloud properties are also analyzed by region and by dynamical regime and thermodynamic conditions. Our results show that for parameters such as total cloud cover, cloud water path, and cloud radiative effect, the CMIP6 multimodel mean performs slightly better than the CMIP5 ensemble mean in terms of mean bias, pattern correlation, and relative root-mean square deviation. The intermodel spread in CMIP6, however, is not reduced compared to CMIP5. Compared with CALIPSO-ICECLOUD data, the CMIP5/6 models overestimate cloud ice, particularly in the lower and middle troposphere, partly due to too high ice fractions for given temperatures. This bias is reduced in the CMIP6 multimodel mean. While many known biases such as an underestimation in cloud cover in stratocumulus regions remain in CMIP6, we find that the CMIP5 problem of too few but too reflective clouds over the Southern Ocean is significantly improved.

Global and Local Assessment of Image Classification Quality on an Overall and Per-Class Basis without Ground Reference Data

Ground reference data are typically required to evaluate the quality of a supervised image classification analysis used to produce a thematic map from remotely sensed data. Acquiring a suitable ground data set for a rigorous assessment of classification quality can be a major challenge. An alternative approach to quality assessment is to use a model-based method such as can be achieved with a latent class analysis. Previous research has shown that the latter can provide estimates of class areal extent for a non-site specific accuracy assessment and yield estimates of producer’s accuracy which are commonly used in site-specific accuracy assessment. Here, the potential for quality assessment via a latent class analysis is extended to show that an estimate of a complete confusion matrix can be predicted which allows a suite of standard accuracy measures to be generated to indicate global quality on an overall and per-class basis. In addition, information on classification uncertainty may be used to illustrate classification quality on a per-pixel basis and hence provide local information to highlight spatial variations in classification quality. Classifications of imagery from airborne and satellite-borne sensors were used to illustrate the potential of the latent class analysis with results compared against those arising from the use of a conventional ground data set.

Spectral mixture analysis for surveillance of harmful algal blooms (SMASH): A field-, laboratory-, and satellite-based approach to identifying cyanobacteria genera from remotely sensed data

Algal blooms around the world are increasing in frequency and severity, often with the possibility of adverse effects on human and ecosystem health. The health and economic impacts associated with harmful algal blooms, or HABs, provide compelling rationale for developing new methods for monitoring these events via remote sensing. Although concentrations of chlorophyll-a and key pigments like phycocyanin are routinely estimated from satellite images and used to infer algal or cyanobacterial cell counts, current methods are unable to provide information on the taxonomic composition of a bloom. This study introduced a new approach capable of differentiating among genera based on their reflectance characteristics: Spectral Mixture Analysis for Surveillance of HABs, or SMASH. The foundation of SMASH is a multiple endmember spectral mixture analysis (MESMA) algorithm that takes a library of cyanobacteria endmembers and a hyperspectral image as input and estimates the fractional abundance of each genus, plus water, on a per-pixel basis. Importantly, we assume that the water column consists of only pure water and cyanobacteria, implying that our linear spectral unmixing models do not account for other optically active constituents such as suspended sediment and colored dissolved organic matter (CDOM). We used reflectance spectra for 12 genera measured under a microscope to populate an algal spectral library and applied the SMASH workflow to satellite images from four waterbodies across the United States. Normalized spectral separability scores indicated that the 12 genera were distinct from one another and the MESMA algorithm reproduced known input fractions for simulated mixtures that included all pairwise combinations of genera and water. We used Upper Klamath Lake as an example to illustrate data products generated via SMASH: maps of the normalized difference chlorophyll index and cyanobacterial index, a MESMA-based classification of algal genera, fraction images for each endmember, and a root mean square error (RMSE) image that summarizes uncertainty. For Upper Klamath Lake, these outputs highlighted a complex algal bloom featuring several genera, primarily Aphanizomenon, and intricate spatial patterns associated with gyres. The maximum RMSE constraint imposed on the MESMA algorithm provided a means of avoiding false positive detection of genera not present in a waterbody but must not be set so low as to leave much of an image unclassified in cases where genera included in the library are present. Comparison of endmember fractions with relative biovolumes calculated from field samples indicated that taxonomic information from SMASH was consistent with field observations. For example, the algorithm successfully identified Microcystis in Owasco Lake but avoided misclassifying Asterionella, a genus not yet included in our library, in Detroit Lake. This proof-of-concept investigation demonstrates the potential of SMASH to enhance our understanding of algal blooms, particularly with respect to their spatial and temporal dynamics.

The influence of vegetation index thresholding on EO-based assessments of exposed soil masks in Germany between 1984 and 2019

Knowledge about the spatial and temporal distribution of exposed soils is necessary for e.g., soil erosion mitigation. Earth Observation (EO) is a valuable data source for detecting exposed soils on a large scale. In the last couple of years, the multitemporal compositing technique has been used for the generation of so-called exposed soil composites that overcome the limitation of temporarily coverage of the soils with vegetation as it is occurring at agricultural sites. The selection of exposed soil pixels from the stack of multispectral images is mainly done using spectral reflectance indices such as NDVI, NBR2 and others calculated on a per-pixel basis. The definition of the thresholds that are applicable to large areas such as regions, countries or continents is still a challenge and requires a reliable and robust sampling data base. In this study, the Soil Composite Mapping Processor (SCMaP) is used to build exposed soil masks containing all pixels in a given time period showing at least once exposed soil. For this purpose, a modified vegetation index (PV) based on the NDVI is used to separate the soils from other land cover (LC) classes by two PV thresholds. The overall goal of this study is to derive and validate exposed soil masks from multi-year Landsat data stacks for Germany from 1984 to 2019. The first focus is set on the impact of a newly developed sampling approach of LC classes such as urban areas, deciduous forests and agricultural fields that are automatically derived from Corine Land Cover (CLC) data. The spectral-temporal behavior of these LC classes in PVmin/max index composites show larger variability of the PV values compared to a manual sampling for selective LC classes such as urban areas. It reveals that the threshold definition method previously developed by Rogge et al. (2018) is not robust enough and the percentile rule used to define the Tmax threshold had to be adapted from 0.995 to 0.900. On the other hand, the sampling data base has proven to be robust across time and region. The second focus of the paper is to validate all generated exposed soil masks covering Germany for seven time periods from 1984 to 2019. A linear correlation analysis was performed comparing the SCMaP data with surveys from the Federal Statistical Office (Destatis) and the CLC inventories. The comparison with both datasets showed high regression coefficients (R2 = 0.79 to 0.90) with small regional deviations for areas in the Northern part of Germany. Strong correlation was found for time periods based on a higher number of cloud free Landsat images such as from 2000 to 2009. This demonstrates the high potential of SCMaP’s to generate exposed soil masks based on an automated sampling and a robust threshold derivation. To contribute to soil erosion studies that need information about where and when soils are bare, accurate exposed soil masks in suitable time periods can be of great value.

Croplands intensify regional and global warming according to satellite observations

Croplands, occupying approximately 13% of the Earth's land surface, can dramatically alter surface temperature via multiple biophysical processes. However, it remains unclear whether and to what extent croplands warm or cool the land surface globally. Here we present new evidence of the temperature effects of global croplands based on high-resolution satellite observations of land surface temperature (LST). We find general daytime warming and nighttime cooling of crop-dominated lands (cropland intensity ≥90%) relative to other vegetation cover, with net daily cooling in the arid zone and net daily warming in all other areas. On a per-pixel basis, however, nearly half of the crop-dominated lands cool daytime LST and/or warm nighttime LST. The temperature effects vary greatly by location and season. The per-pixel intensity can exceed 10 °C likely due to the changes in surface evapotranspiration, albedo, and background climate. By extending the LST changes of crop-dominated lands to the full land surface, we find that croplands make large contributions to regional and global land surface warming. Daytime warming can reach 0.56 °C in Europe during spring. Nationally, China, India, and Brazil lead the warming due to their large cropland areas and strong local climate effects. Our results highlight the importance of accounting for the biophysical effects of croplands for projecting climate change and formulating adaptation and mitigation strategies.

Automated detection of aggressive and indolent prostate cancer on magnetic resonance imaging.

PurposeWhile multi‐parametric magnetic resonance imaging (MRI) shows great promise in assisting with prostate cancer diagnosis and localization, subtle differences in appearance between cancer and normal tissue lead to many false positive and false negative interpretations by radiologists. We sought to automatically detect aggressive cancer (Gleason pattern ≥ 4) and indolent cancer (Gleason pattern 3) on a per‐pixel basis on MRI to facilitate the targeting of aggressive cancer during biopsy.MethodsWe created the Stanford Prostate Cancer Network (SPCNet), a convolutional neural network model, trained to distinguish between aggressive cancer, indolent cancer, and normal tissue on MRI. Ground truth cancer labels were obtained by registering MRI with whole‐mount digital histopathology images from patients who underwent radical prostatectomy. Before registration, these histopathology images were automatically annotated to show Gleason patterns on a per‐pixel basis. The model was trained on data from 78 patients who underwent radical prostatectomy and 24 patients without prostate cancer. The model was evaluated on a pixel and lesion level in 322 patients, including six patients with normal MRI and no cancer, 23 patients who underwent radical prostatectomy, and 293 patients who underwent biopsy. Moreover, we assessed the ability of our model to detect clinically significant cancer (lesions with an aggressive component) and compared it to the performance of radiologists.ResultsOur model detected clinically significant lesions with an area under the receiver operator characteristics curve of 0.75 for radical prostatectomy patients and 0.80 for biopsy patients. Moreover, the model detected up to 18% of lesions missed by radiologists, and overall had a sensitivity and specificity that approached that of radiologists in detecting clinically significant cancer.ConclusionsOur SPCNet model accurately detected aggressive prostate cancer. Its performance approached that of radiologists, and it helped identify lesions otherwise missed by radiologists. Our model has the potential to assist physicians in specifically targeting the aggressive component of prostate cancers during biopsy or focal treatment.