Search

Detecting small objects (e.g., manhole covers, license plates, and roadside milestones) in urban images is a long-standing challenge mainly due to the scale of small object and background clutter. Although convolution neural network (CNN)-based methods have made significant progress and achieved impressive results in generic object detection, the problem of small object detection remains unsolved. To address this challenge, in this study we developed an end-to-end network architecture that has three significant characteristics compared to previous works. First, we designed a backbone network module, namely Reduced Downsampling Network (RD-Net), to extract informative feature representations with high spatial resolutions and preserve local information for small objects. Second, we introduced an Adjustable Sample Selection (ADSS) module which frees the Intersection-over-Union (IoU) threshold hyperparameters and defines positive and negative training samples based on statistical characteristics between generated anchors and ground reference bounding boxes. Third, we incorporated the generalized Intersection-over-Union (GIoU) loss for bounding box regression, which efficiently bridges the gap between distance-based optimization loss and area-based evaluation metrics. We demonstrated the effectiveness of our method by performing extensive experiments on the public Urban Element Detection (UED) dataset acquired by Mobile Mapping Systems (MMS). The Average Precision (AP) of the proposed method was 81.71%, representing an improvement of 1.2% compared with the popular detection framework Faster R-CNN.

Extensive, complex supraglacial river networks form on the southwest Greenland ice sheet (GrIS) surface each melt season. These networks are the dominant pathways for surface meltwater transport on this part of the ice sheet, but their fluvial morphometry has received little study. This paper utilizes high-resolution (2 m) WorldView-1/2 images, digital elevation models, and GIS tools to present a detailed morphometric characterization (river number, river length, Strahler stream order, width, depth, bifurcation ratio, braiding index, drainage density, slope, and relief ratio) for 523 GrIS supraglacial river networks. A new algorithm is presented to determine Strahler stream order in supraglacial environments. Results show that (1) Supraglacial river networks are broadly similar to terrestrial landscapes in that they follow Horton’s laws (river number, mean river length, and slope versus stream order), widen downstream, and have comparable mean bifurcation ratios (3.7 ± 1.9) and braiding indices; (2) unlik...

Meltwater runoff from the Greenland ice sheet surface influences surface mass balance (SMB), ice dynamics, and global sea level rise, but is estimated with climate models and thus difficult to validate. We present a way to measure ice surface runoff directly, from hourly in situ supraglacial river discharge measurements and simultaneous high-resolution satellite/drone remote sensing of upstream fluvial catchment area. A first 72-h trial for a 63.1-km

Significance Meltwater runoff is an important hydrological process operating on the Greenland ice sheet surface that is rarely studied directly. By combining satellite and drone remote sensing with continuous field measurements of discharge in a large supraglacial river, we obtained 72 h of runoff observations suitable for comparison with climate model predictions. The field observations quantify how a large, fluvial supraglacial catchment attenuates the magnitude and timing of runoff delivered to its terminal moulin and hence the bed. The data are used to calibrate classical fluvial hydrology equations to improve meltwater runoff models and to demonstrate that broad-scale surface water drainage patterns that form on the ice surface powerfully alter the timing, magnitude, and locations of meltwater penetrating into the ice sheet., Meltwater runoff from the Greenland ice sheet surface influences surface mass balance (SMB), ice dynamics, and global sea level rise, but is estimated with climate models and thus difficult to validate. We present a way to measure ice surface runoff directly, from hourly in situ supraglacial river discharge measurements and simultaneous high-resolution satellite/drone remote sensing of upstream fluvial catchment area. A first 72-h trial for a 63.1-km2 moulin-terminating internally drained catchment (IDC) on Greenland’s midelevation (1,207–1,381 m above sea level) ablation zone is compared with melt and runoff simulations from HIRHAM5, MAR3.6, RACMO2.3, MERRA-2, and SEB climate/SMB models. Current models cannot reproduce peak discharges or timing of runoff entering moulins but are improved using synthetic unit hydrograph (SUH) theory. Retroactive SUH applications to two older field studies reproduce their findings, signifying that remotely sensed IDC area, shape, and supraglacial river length are useful for predicting delays in peak runoff delivery to moulins. Applying SUH to HIRHAM5, MAR3.6, and RACMO2.3 gridded melt products for 799 surrounding IDCs suggests their terminal moulins receive lower peak discharges, less diurnal variability, and asynchronous runoff timing relative to climate/SMB model output alone. Conversely, large IDCs produce high moulin discharges, even at high elevations where melt rates are low. During this particular field experiment, models overestimated runoff by +21 to +58%, linked to overestimated surface ablation and possible meltwater retention in bare, porous, low-density ice. Direct measurements of ice surface runoff will improve climate/SMB models, and incorporating remotely sensed IDCs will aid coupling of SMB with ice dynamics and subglacial systems.

Digital elevation models (DEMs) of ice surface topography are often used for hydrologic analysis of the Greenland ice sheet (GrIS), but their suitability for this purpose has received little quantitative assessment. We compare remotely sensed maps of supraglacial lakes, rivers, and moulins with their DEM-modeled counterparts, using two moderate-resolution DEMs (SPIRIT DEM and ASTER GDEM) for a ${\sim}24\,000\;\text{km}^{2}$ area of the southwestern GrIS. We find that modeled hydrological features are critically sensitive to selection of a depression area threshold (a user-specified parameter used to fill noise and/or true topographic depressions in the ice surface DEM), with small depression area thresholds over-predicting observed supraglacial lake abundance and large thresholds under-predicting lake abundance. Few remotely sensed moulins are identified in either DEM, even if a small depression area threshold is used. A standard practice of filling all DEM depressions yields modeled surface flow paths that broadly match remotely sensed supraglacial river networks, but are far less fragmented than reality (owing to moulin capture), raising into question the realism of this standard practice. In sum, moderate-resolution DEMs do hold value for simulating broad-scale hydrography of the GrIS surface, but are critically sensitive to choice of the filling threshold, and insensitive to moulins which also influence supraglacial drainage pattern. Our preliminary analysis suggests using a depression area threshold of $0.1{-}0.2\;\text{km}^{2}$ for lakes, advises against using DEMs to predict moulin locations, and urges caution when using 100% DEM filling to model flow paths of supraglacial rivers.

The presence of SARS-CoV-2 was officially documented in Europe at the end of February 2020. Despite many observations, the real impact of COVID-19 in the European Union (EU), its underlying factors and their contribution to mortality and morbidity outcomes were never systematically investigated. The aim of the present work is to provide an overview and a meta-analysis of main predictors and of country differences of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection-associated mortality rate (MR) in hospitalized patients. Out of 3714 retrieved articles, 87 studies were considered, including 35,486 patients (mean age 60.9 ± 8.2 years) and 5867 deaths. After adjustment for confounders, diabetes mellitus was the best predictors of MR in an age- and sex-dependent manner, followed by chronic pulmonary obstructive diseases and malignancies. In both the US and Europe, MR was higher than that reported in Asia (25[20;29] % and 20[17;23] % vs. 13[10;17]%; both p