Automatic Detection of Ditches and Natural Streams from Digital Elevation Models Using Deep Learning

Computers & Geosciences, Год журнала: 2025, Номер unknown, С. 105875 - 105875

Опубликована: Янв. 1, 2025

Язык: Английский

---

Elevation-dependent patterns of temporally asymmetrical vegetation response to climate in an alpine basin on the Qinghai-Tibet Plateau

Ecological Indicators, Год журнала: 2024, Номер 159, С. 111736 - 111736

Опубликована: Фев. 1, 2024

The temporally asymmetric effects of vegetation growth response to climate, e.g., time-lag, time-accumulation, and their combination effects, have been widely reported. However, elevation-dependent spatial distribution patterns in alpine basins the Qinghai-Tibet Plateau remain unclear, especially different seasons. This study investigated temporal precipitation temperature throughout whole growing season its two subphases (growth phase senescence phase) Lhasa River Basin (LRB). results showed that responds climate all seasons with obvious time-accumulation relatively weak during season, there is no significant lag both temperature, while accumulation about 3 ∼ 4 weeks 2 weeks, respectively. Comparatively, are enhanced phase, a similar 6-week 4-week for LRB subdivided into four elevation zones, each hydrothermal condition characteristic. Across it found stronger water-deficit stress on shorter longer precipitation, heat-deficit leads which determined by vegetation's demands ambient environment. These findings shed light relationship between provide potential theoretical support conservation basins.

Язык: Английский

---

Machine Learning and Deep Learning in Remote Sensing Data Analysis

Elsevier eBooks, Год журнала: 2024, Номер unknown

Опубликована: Янв. 1, 2024

Язык: Английский

---

DeepWaterFraction: A globally applicable, self-training deep learning approach for percent surface water area estimation from Landsat mission imagery

Journal of Hydrology, Год журнала: 2024, Номер 638, С. 131512 - 131512

Опубликована: Июнь 15, 2024

Язык: Английский

---

Reply on RC1

Опубликована: Янв. 16, 2025

Abstract. Rivers play important roles in ecological biodiversity, shipping trade and the carbon cycle. Owing to human disturbances extreme climates recent decades, river extents have altered frequently dramatically. The development of sequential fine-scale extent datasets, which could offer strong data support for protection, management sustainable use, is urgently needed. A literature review revealed that annual datasets with fine spatial resolutions are generally unavailable China. To address this issue, first Sentinel-derived China dataset (CRED) from 2016 2023 was produced our study. We water maps by combining dynamic world (DW), ESRI global land cover (EGLC) multiple index detection rule (MIWDR). For DW MIWDR time series, mode algorithm, calculates most common values, used generate yearly maps. Then, an object-based hierarchical decision tree based on geometric features auxiliary developed extract rivers data. results indicated overall accuracies (OAs) CRED were greater than 96.0 % 2023. user (UAs), producer (PAs) F1 scores exceeded 95.3 %, 91.3 93.7 respectively. further intercomparison shared similar patterns wetland map East Asia (EA_Wetlands), use/cover change (CNLUCC) covers (CWaC) correlation coefficients (R) 0.75. Moreover, outperformed three terms small mapping misclassification reduction. area statistics 44,948.78 km² 2023, mostly distributed coastal provinces From areas characterized initial increase, followed a decrease then slight increase. Spatially, decreased located mainly Southeast China, whereas increased Central Northeast In general, explicitly delineated dynamics provide good foundation improving ecology management. publicly available at https://doi.org/10.5281/zenodo.13841910 (Peng et al., 2024a).

Язык: Английский

---

Reply on RC2

Опубликована: Янв. 16, 2025

Abstract. Rivers play important roles in ecological biodiversity, shipping trade and the carbon cycle. Owing to human disturbances extreme climates recent decades, river extents have altered frequently dramatically. The development of sequential fine-scale extent datasets, which could offer strong data support for protection, management sustainable use, is urgently needed. A literature review revealed that annual datasets with fine spatial resolutions are generally unavailable China. To address this issue, first Sentinel-derived China dataset (CRED) from 2016 2023 was produced our study. We water maps by combining dynamic world (DW), ESRI global land cover (EGLC) multiple index detection rule (MIWDR). For DW MIWDR time series, mode algorithm, calculates most common values, used generate yearly maps. Then, an object-based hierarchical decision tree based on geometric features auxiliary developed extract rivers data. results indicated overall accuracies (OAs) CRED were greater than 96.0 % 2023. user (UAs), producer (PAs) F1 scores exceeded 95.3 %, 91.3 93.7 respectively. further intercomparison shared similar patterns wetland map East Asia (EA_Wetlands), use/cover change (CNLUCC) covers (CWaC) correlation coefficients (R) 0.75. Moreover, outperformed three terms small mapping misclassification reduction. area statistics 44,948.78 km² 2023, mostly distributed coastal provinces From areas characterized initial increase, followed a decrease then slight increase. Spatially, decreased located mainly Southeast China, whereas increased Central Northeast In general, explicitly delineated dynamics provide good foundation improving ecology management. publicly available at https://doi.org/10.5281/zenodo.13841910 (Peng et al., 2024a).

Язык: Английский

---

Enhanced River Connectivity Assessment Across Larger Areas Through Deep Learning With Dam Detection

Hydrological Processes, Год журнала: 2025, Номер 39(1)

Опубликована: Янв. 1, 2025

ABSTRACT Monitoring river connectivity across large regions is essential for understanding hydrological processes and environmental management. However, comprehensive assessments of are often hindered by inaccurate dam databases, which biased towards larger dams while overlooking smaller or low‐head dams. To enhance the accuracy assessments, we developed three advanced convolutional neural networks (CNNs; YOLOv5, Advance‐You Only Look Once [YOLO], Faster R‐CNN) to accurately classify evaluate using high‐resolution (1 m) remote sensing imagery. The evaluation results showed that Advance‐YOLO performs best with an average mean precision (mAP) 86.6%, R‐CNN mediocrely mAP 77.9%. Applying well‐trained model in Tarim River Basin (China), one largest inland basins around globe, found there currently 135 total on its sources. Conversely, existing public database underestimates 85.9% Notably, a 14.3% decline over past decade, current density four source rivers 1.12 per 10 000 km 2 . overestimated 83.9%. here enhances assessment areas long period, thereby fostering more research effective water resource

Язык: Английский

---

Predicting the thickness of alpine meadow soil on headwater hillslopes of the Qinghai-Tibet Plateau

Geoderma, Год журнала: 2025, Номер 456, С. 117271 - 117271

Опубликована: Март 26, 2025

Язык: Английский

---

Characteristics of the water extent and width of endorheic Tibetan Plateau rivers revealed by Sentinel-2

Journal of Hydrology, Год журнала: 2025, Номер unknown, С. 133191 - 133191

Опубликована: Март 1, 2025

Язык: Английский

---

Extracting an accurate river network: Stream burning re-revisited

Remote Sensing of Environment, Год журнала: 2024, Номер 312, С. 114333 - 114333

Опубликована: Июль 29, 2024

Extracting river networks that are both accurate and topologically connected is important for applications involve correct routing of material, example water sediment, through such networks. We combined sediment extraction using radar multispectral imagery from Sentinel-1 Sentinel-2 to create masks over a range study areas. These were then used condition topographic Digital Elevation Models (DEMs) by lowering the elevation pixels with present, in process known as stream burning. examined how burning could improve accuracy extracted identified most effective method optimal results. find deeper depths improved accuracy, diminishing returns: we suggest 40 50 meters. improves humid temperate landscapes, but arid landscapes should be burned only pixels. significantly better on COP30 global dataset compared NASADEM dataset, mainly due time collection. The AW3D30 DEM FABDEM datasets have accuracies just below DEM.

Язык: Английский

---