Reinhard Hinkelmann
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Reinhard Hinkelmann has a total of 341 co-authors. Here are the top co-authors by number of shared publications:
- Ilhan Özgen-Xian
IGÖ, Technische Universität Braunschweig, Langer Kamp 19c, 38106 Brunswick, Germany
21 shared publications Jingming Hou
State Key Laboratory of Eco-hydraulics in Northwest Arid Region of China, Xi’an University of Technology, Xi’an, China
15 shared publications- Franziska Tügel
University of Twente
8 shared publications - Jingming Hou
State Key Laboratory of Water Engineering Ecology and Environment in Arid Area, Xi’an University of Technology, Xi’an, China
8 shared publications - Mohammad Zounemat-Kermani
Department of Civil Engineering, Shahid Bahonar University of Kerman, Kerman, Iran
8 shared publications Qiuhua Liang
8 shared publications
Latest Publications
- Open Access
- 19 November 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (22) , 3976
Decoupling Ecological Trends from Stochastic Climate Volatility: Spatiotemporal Transitions and Future Resilience in Arid and Semi-Arid Regions
Arid ecosystems face severe climate threats, yet stochastic climate volatility often masks true ecological transitions. In this study, we developed an integrated spatiotemporal framework to decouple mean ecological trends from climate noise in Gansu Province, China. We combined Remote Sensing Ecological Index (RSEI) time series, the multi-scalar Standardized Precipitation Evapotranspiration Index (SPEI), and a probabilistic Dust Hazard Index (DHI) to evaluate historical dynamics (2001–2024). Subsequently, eco-environmental sustainability (2025–2099) was projected under CMIP6 scenarios using a Random Forest algorithm optimized via spatial block cross-validation. Historical evaluations revealed a significant regional ecological amelioration (RSEI trend: +0.0279/decade, p < 0.01), with the degraded area decreasing from 68.60% in 2001 to 52.68% in 2024. However, spatial polarization persisted, with severe degradation covering 38.84% of the hyper-arid northwest. Feature importance and dual correlation analyses identified short-term moisture deficits (SPEI-3) and chronic dust exposure as primary biophysical controls; notably, dust hazards strongly suppressed vegetation in degraded zones (r = −0.43). Future projections indicate that warming-induced aridity will intensify ecological stress. Under the intermediate SSP2-4.5 scenario, the area of Serious Decline (SSD) expands from 35.09% in the near term to 47.15% by the end of the century. Conversely, the high-emission SSP5-8.5 scenario, characterized by thermal extremes up to 3.58 °C, triggers an abrupt structural transition during 2061–2099; the SSD area more than triples (from 12.13% to 43.95%), overwhelming historical restoration gains. Safeguarding these fragile drylands necessitates proactive management integrating targeted dust mitigation and water rights allocation.
- 19 November 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (22) , 3976
Open AccessDecoupling Ecological Trends from Stochastic Climate Volatility: Spatiotemporal Transitions and Future Resilience in Arid and Semi-Arid Regions
Arid ecosystems face severe climate threats, yet stochastic climate volatility often masks true ecological transitions. In this study, we developed an integrated spatiotemporal framework to decouple mean ecological trends from climate noise in Gansu Province, China. We combined Remote Sensing Ecological Index (RSEI) time series, the multi-scalar Standardized Precipitation Evapotranspiration Index (SPEI), and a probabilistic Dust Hazard Index (DHI) to evaluate historical dynamics (2001–2024). Subsequently, eco-environmental sustainability (2025–2099) was projected under CMIP6 scenarios using a Random Forest algorithm optimized via spatial block cross-validation. Historical evaluations revealed a significant regional ecological amelioration (RSEI trend: +0.0279/decade, p < 0.01), with the degraded area decreasing from 68.60% in 2001 to 52.68% in 2024. However, spatial polarization persisted, with severe degradation covering 38.84% of the hyper-arid northwest. Feature importance and dual correlation analyses identified short-term moisture deficits (SPEI-3) and chronic dust exposure as primary biophysical controls; notably, dust hazards strongly suppressed vegetation in degraded zones (r = −0.43). Future projections indicate that warming-induced aridity will intensify ecological stress. Under the intermediate SSP2-4.5 scenario, the area of Serious Decline (SSD) expands from 35.09% in the near term to 47.15% by the end of the century. Conversely, the high-emission SSP5-8.5 scenario, characterized by thermal extremes up to 3.58 °C, triggers an abrupt structural transition during 2061–2099; the SSD area more than triples (from 12.13% to 43.95%), overwhelming historical restoration gains. Safeguarding these fragile drylands necessitates proactive management integrating targeted dust mitigation and water rights allocation.
- Open Access
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
Resolution–Acceptance and Uncertainty Trade-offs between SonTek RS5 Auto and User-Defined High-Resolution Profiling Modes in an Open-Channel Flume
Acoustic Doppler current profilers (ADCPs) are widely used in open-channel flows, but their performance depends on processing configuration in ways not well characterized. This study evaluates the SonTek RS5 under four configurations combining Auto mode (cell size, cz = 0.050 m) and a user-defined high-resolution Advance mode (cz = 0.025 m) with clear-water and sediment-laden conditions across six Reynolds numbers (Re = 6.57×104–1.79×105) in a flume with two-dimensional two-component particle image velocimetry (2D2C-PIV) as reference. Because the modes differ in cell start distance and acoustic sub-mode, results reflect configuration effects rather than cell size alone. Advance mode extends the profile from y/H = 0.15–0.69 to 0.14–0.81 (y/H: normalized height) and reduces near-bed profile instabilities, while inter-mode core deviation reaches 9.8%. In clear water, the median fraction of pings passing the correlation filter (%good) decreases from 0.99–1.00 in Auto mode to 0.71–0.96 in Advance mode, a coverage–acceptance trade-off. Under the tested sand-bed condition, this trade-off is reduced: Advance %good rises to 0.96–1.00, inter-mode core deviation is at most 3.7%, and near-bed profiles appear more stable, although bed material and acoustic scattering changed together. Auto mode mixes pulse-coherent, broadband, and hybrid sub-modes at intermediate and high Re; Advance mode remains pulse-coherent. Expanded core-band uncertainty is 0.0038–0.0083 m/s. Against 2D2C-PIV, Advance mode yields relative RMSE of 0.71–5.20%. These results suggest that the previously reported Auto mode overestimation of up to 32% was at least partly configuration-dependent rather than intrinsic to the ADCP measurement principle, and provide practical guidance for RS5 processing-mode selection.
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessResolution–Acceptance and Uncertainty Trade-offs between SonTek RS5 Auto and User-Defined High-Resolution Profiling Modes in an Open-Channel Flume
Acoustic Doppler current profilers (ADCPs) are widely used in open-channel flows, but their performance depends on processing configuration in ways not well characterized. This study evaluates the SonTek RS5 under four configurations combining Auto mode (cell size, cz = 0.050 m) and a user-defined high-resolution Advance mode (cz = 0.025 m) with clear-water and sediment-laden conditions across six Reynolds numbers (Re = 6.57×104–1.79×105) in a flume with two-dimensional two-component particle image velocimetry (2D2C-PIV) as reference. Because the modes differ in cell start distance and acoustic sub-mode, results reflect configuration effects rather than cell size alone. Advance mode extends the profile from y/H = 0.15–0.69 to 0.14–0.81 (y/H: normalized height) and reduces near-bed profile instabilities, while inter-mode core deviation reaches 9.8%. In clear water, the median fraction of pings passing the correlation filter (%good) decreases from 0.99–1.00 in Auto mode to 0.71–0.96 in Advance mode, a coverage–acceptance trade-off. Under the tested sand-bed condition, this trade-off is reduced: Advance %good rises to 0.96–1.00, inter-mode core deviation is at most 3.7%, and near-bed profiles appear more stable, although bed material and acoustic scattering changed together. Auto mode mixes pulse-coherent, broadband, and hybrid sub-modes at intermediate and high Re; Advance mode remains pulse-coherent. Expanded core-band uncertainty is 0.0038–0.0083 m/s. Against 2D2C-PIV, Advance mode yields relative RMSE of 0.71–5.20%. These results suggest that the previously reported Auto mode overestimation of up to 32% was at least partly configuration-dependent rather than intrinsic to the ADCP measurement principle, and provide practical guidance for RS5 processing-mode selection.
- Open Access
- 4 September 2026
- Published by Taylor & Francis in Digital Water
- Vol. 4 (1)
High-performance computing in the hydrosciences
1. Within the last decades, High Performance Computing (HPC) has evolved into the third pillar in the hydrosciences, in addition to theory and experiments. It is characterised by tremendous increas...
- 4 September 2026
- Published by Taylor & Francis in Digital Water
- Vol. 4 (1)
Open AccessHigh-performance computing in the hydrosciences
1. Within the last decades, High Performance Computing (HPC) has evolved into the third pillar in the hydrosciences, in addition to theory and experiments. It is characterised by tremendous increas...
- Open Access
- 24 August 2026
- Published by Elsevier in SSRN Electronic Journal
Full-Depth Velocity Characterization in Open-Channel Flow Using Complementary ADCP, PIV, and CFD: A Regional Framework Across Reynolds Numbers
Full-depth velocity profile characterization in open-channel flow is often limited by the spatial coverage of individual measurement methods. Acoustic Doppler Current Profiler (ADCP) measurements are limited near the free surface by transducer submergence and near-transducer blanking distance, and near the bed by bottom interference and side-lobe contamination from the bed echo, while Particle Image Velocimetry (PIV) is constrained by optical access. Computational Fluid Dynamics (CFD) can provide full-depth continuity, but its use in unsampled regions requires experimental validation. This study presents a three-method benchmark combining 2D2C-PIV, ADCP, and CFD across six Reynolds numbers (Re = 6.48 × 10⁴ to 1.76 × 10⁵) in a rectangular open channel.A regional framework defines Region I (near-bed blind zone, y/H < 0.141), Region II (three-method overlap, 0.141 ≤ y/H ≤ 0.467), Region III (upper flow, 0.467 < y/H ≤ 0.810), and Region IV (near-surface blind zone, y/H > 0.810). CFD was validated against PIV in Regions I–II with relative RMSE ≤ 4.51 %. In Region II, PIV and ADCP agreed most closely, with relative RMSE below 1.3 % for four of the six cases. In Region III, ADCP exceeded CFD by 7.6–10.7 %; without an independent reference, the offset cannot be attributed to either method alone. Blind-zone analysis shows that Region I contains the strongest velocity gradients, with velocity rising from zero to 94–97 % of the depth-mean value by y/H ≈ 0.141, whereas Region IV is nearly uniform. The combined approach provides a practical framework for velocity-profile characterization and ADCP blind-zone interpretation.
- 24 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessFull-Depth Velocity Characterization in Open-Channel Flow Using Complementary ADCP, PIV, and CFD: A Regional Framework Across Reynolds Numbers
Full-depth velocity profile characterization in open-channel flow is often limited by the spatial coverage of individual measurement methods. Acoustic Doppler Current Profiler (ADCP) measurements are limited near the free surface by transducer submergence and near-transducer blanking distance, and near the bed by bottom interference and side-lobe contamination from the bed echo, while Particle Image Velocimetry (PIV) is constrained by optical access. Computational Fluid Dynamics (CFD) can provide full-depth continuity, but its use in unsampled regions requires experimental validation. This study presents a three-method benchmark combining 2D2C-PIV, ADCP, and CFD across six Reynolds numbers (Re = 6.48 × 10⁴ to 1.76 × 10⁵) in a rectangular open channel.A regional framework defines Region I (near-bed blind zone, y/H < 0.141), Region II (three-method overlap, 0.141 ≤ y/H ≤ 0.467), Region III (upper flow, 0.467 < y/H ≤ 0.810), and Region IV (near-surface blind zone, y/H > 0.810). CFD was validated against PIV in Regions I–II with relative RMSE ≤ 4.51 %. In Region II, PIV and ADCP agreed most closely, with relative RMSE below 1.3 % for four of the six cases. In Region III, ADCP exceeded CFD by 7.6–10.7 %; without an independent reference, the offset cannot be attributed to either method alone. Blind-zone analysis shows that Region I contains the strongest velocity gradients, with velocity rising from zero to 94–97 % of the depth-mean value by y/H ≈ 0.141, whereas Region IV is nearly uniform. The combined approach provides a practical framework for velocity-profile characterization and ADCP blind-zone interpretation.
- Open Access
- 6 August 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (15) , 2630
Spatial Heterogeneity and Driving Mechanisms of Carbon Storage on the Chinese Loess Plateau
The Chinese Loess Plateau is a vast semi-arid landscape that has experienced substantial vegetation recovery through afforestation since the late twentieth century, altering the regional carbon cycle and contributing to China’s carbon neutrality efforts. However, large-scale quantification of carbon storage (CS) remains challenging in this heterogeneous and erosion-prone landscape. In this study, the PLUS land-use simulation model was coupled with an InVEST-based carbon storage accounting framework, and multi-source remote sensing and field survey data were integrated to quantify historical carbon storage from 1980 to 2020 and predict carbon storage in 2040 under three future land-use scenarios: natural development (ND), cultivated land protection (CLP), and ecological protection (EP). The results showed that NDVI exhibited a distinct southeast-to-northwest gradient, with peak values (>0.8) in the southeastern tablelands, where CS reached 10.27 ± 4.20 kg C m−2. Carbon storage increased from 1980 to 2020, primarily due to afforestation, with higher CS in the humid southeastern plateau and lower CS in the arid northwest. Scenario simulation showed that CS maintained a baseline level of 7.07 ± 3.81 kg C m−2 under the ND scenario, increased slightly to 7.08 ± 3.77 kg C m−2 under the CLP scenario, and reached the highest value of 7.10 ± 3.80 kg C m−2 under the EP scenario. SEM-based mechanism analysis indicated that topography exerted the strongest direct influence on CS, followed by vegetation, climate, and socioeconomic factors. These findings suggest that water availability and topographic constraints strongly regulate CS in semi-arid regions, while land-use policy can still modify CS through scenario-dependent land-use change.
- 6 August 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (15) , 2630
Open AccessSpatial Heterogeneity and Driving Mechanisms of Carbon Storage on the Chinese Loess Plateau
The Chinese Loess Plateau is a vast semi-arid landscape that has experienced substantial vegetation recovery through afforestation since the late twentieth century, altering the regional carbon cycle and contributing to China’s carbon neutrality efforts. However, large-scale quantification of carbon storage (CS) remains challenging in this heterogeneous and erosion-prone landscape. In this study, the PLUS land-use simulation model was coupled with an InVEST-based carbon storage accounting framework, and multi-source remote sensing and field survey data were integrated to quantify historical carbon storage from 1980 to 2020 and predict carbon storage in 2040 under three future land-use scenarios: natural development (ND), cultivated land protection (CLP), and ecological protection (EP). The results showed that NDVI exhibited a distinct southeast-to-northwest gradient, with peak values (>0.8) in the southeastern tablelands, where CS reached 10.27 ± 4.20 kg C m−2. Carbon storage increased from 1980 to 2020, primarily due to afforestation, with higher CS in the humid southeastern plateau and lower CS in the arid northwest. Scenario simulation showed that CS maintained a baseline level of 7.07 ± 3.81 kg C m−2 under the ND scenario, increased slightly to 7.08 ± 3.77 kg C m−2 under the CLP scenario, and reached the highest value of 7.10 ± 3.80 kg C m−2 under the EP scenario. SEM-based mechanism analysis indicated that topography exerted the strongest direct influence on CS, followed by vegetation, climate, and socioeconomic factors. These findings suggest that water availability and topographic constraints strongly regulate CS in semi-arid regions, while land-use policy can still modify CS through scenario-dependent land-use change.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Agent-Based Dynamic Vulnerability Model for Pedestrians Exposed to Floodwaters in Critical Infrastructures
Climate change-amplified flooding poses severe risks to urban underground infrastructures, increasing exposure and vulnerability in densely populated cities. Motivated by the observation that current assessment methods may underestimate the impact of human motions in floodwaters on pedestrian evacuation safety, while traditional evacuation designs primarily focus on individual behavior, neglecting the critical influence of group dynamics and collective decision-making during real flood events. To address these gaps, this study develops an agent-based dynamic vulnerability model for pedestrians exposed to floodwaters, supported by a full-scale instrumented physical model to capture interactive and dynamic evacuation behaviors. The model incorporates group interactions, formation patterns, and hydrodynamic forces acting on pedestrians during evacuation. Analysis of spatial and temporal dynamics of pedestrian movement reveals significant variations in stability: walking against the flow increases instability and overall vulnerability, whereas moving with the flow reduces hydrodynamic forces, though this effect diminishes with increasing water depth. Preliminary results also indicate that group dynamics significantly influence evacuation efficiency: larger spacing between pedestrians mitigates hydrodynamic impacts and enhances evacuation performance, while lateral formations experience higher hydrodynamic forces compared with longitudinal formations, reducing overall efficiency. Integration of the multi-agent model into a hydrodynamic simulation framework enables comprehensive risk assessment and management of underground infrastructure under extreme flooding, facilitating identification of optimal evacuation timing and routing strategies. This framework provides practical guidance for designing flood-resilient underground spaces and contributes a novel approach for dynamic vulnerability assessment in climate-adaptive cities.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessAgent-Based Dynamic Vulnerability Model for Pedestrians Exposed to Floodwaters in Critical Infrastructures
Climate change-amplified flooding poses severe risks to urban underground infrastructures, increasing exposure and vulnerability in densely populated cities. Motivated by the observation that current assessment methods may underestimate the impact of human motions in floodwaters on pedestrian evacuation safety, while traditional evacuation designs primarily focus on individual behavior, neglecting the critical influence of group dynamics and collective decision-making during real flood events. To address these gaps, this study develops an agent-based dynamic vulnerability model for pedestrians exposed to floodwaters, supported by a full-scale instrumented physical model to capture interactive and dynamic evacuation behaviors. The model incorporates group interactions, formation patterns, and hydrodynamic forces acting on pedestrians during evacuation. Analysis of spatial and temporal dynamics of pedestrian movement reveals significant variations in stability: walking against the flow increases instability and overall vulnerability, whereas moving with the flow reduces hydrodynamic forces, though this effect diminishes with increasing water depth. Preliminary results also indicate that group dynamics significantly influence evacuation efficiency: larger spacing between pedestrians mitigates hydrodynamic impacts and enhances evacuation performance, while lateral formations experience higher hydrodynamic forces compared with longitudinal formations, reducing overall efficiency. Integration of the multi-agent model into a hydrodynamic simulation framework enables comprehensive risk assessment and management of underground infrastructure under extreme flooding, facilitating identification of optimal evacuation timing and routing strategies. This framework provides practical guidance for designing flood-resilient underground spaces and contributes a novel approach for dynamic vulnerability assessment in climate-adaptive cities.
- Open Access
- 2 December 2025
- Published by Springer Nature in Arabian Journal of Geosciences
- Vol. 18 (12) , 236
Impact of deforestation on runoff dynamics in the Wardha River sub-basin: a decadal analysis and future projections
Deforestation and urban expansion substantially alter hydrological processes by accelerating runoff, reducing groundwater recharge, and increasing flood risks. The current study assessed runoff dynamics in the Wardha River sub-basin from 2010 to 2020 using the NRCS-CN method, with land-use/land-cover (LULC) and rainfall inputs processed on the Google Earth Engine (GEE) platform and validated against observed discharge. During this period, mean runoff increased by 11.34%, linked to a 23.34% decline in forest cover and a 148% rise in impervious surfaces. Elasticity analysis indicated that rainfall influenced runoff (elasticity > 1), but land-use transitions had a stronger effect, with the 2010–2020 changes showing elasticity 2.07. A hypothetical-future scenario (HFS), simulating full forest-to-cropland and fallow-to-built-up conversion, projected an additional 10% rise in runoff with elasticity 1.91, underscoring the hydrological risks of continued land transformation. Model evaluation confirmed good performance, with R² > 0.90 across scenarios and the HFS scenario showing the best fit (NSE = 0.504, PBIAS = − 29.4%, RMSE = 21.59 mm). Despite scale mismatches between area-averaged runoff and point-based discharge, simulations based on 2010, 2020, and CN-adjusted datasets achieved moderate to acceptable accuracy. These findings highlight the dominant role of land-use change, particularly deforestation, in shaping runoff dynamics and flood risks in the Wardha River sub-basin, compared with the broader but less intense influence of rainfall variability.
- 2 December 2025
- Published by Springer Nature in Arabian Journal of Geosciences
- Vol. 18 (12) , 236
Open AccessImpact of deforestation on runoff dynamics in the Wardha River sub-basin: a decadal analysis and future projections
Deforestation and urban expansion substantially alter hydrological processes by accelerating runoff, reducing groundwater recharge, and increasing flood risks. The current study assessed runoff dynamics in the Wardha River sub-basin from 2010 to 2020 using the NRCS-CN method, with land-use/land-cover (LULC) and rainfall inputs processed on the Google Earth Engine (GEE) platform and validated against observed discharge. During this period, mean runoff increased by 11.34%, linked to a 23.34% decline in forest cover and a 148% rise in impervious surfaces. Elasticity analysis indicated that rainfall influenced runoff (elasticity > 1), but land-use transitions had a stronger effect, with the 2010–2020 changes showing elasticity 2.07. A hypothetical-future scenario (HFS), simulating full forest-to-cropland and fallow-to-built-up conversion, projected an additional 10% rise in runoff with elasticity 1.91, underscoring the hydrological risks of continued land transformation. Model evaluation confirmed good performance, with R² > 0.90 across scenarios and the HFS scenario showing the best fit (NSE = 0.504, PBIAS = − 29.4%, RMSE = 21.59 mm). Despite scale mismatches between area-averaged runoff and point-based discharge, simulations based on 2010, 2020, and CN-adjusted datasets achieved moderate to acceptable accuracy. These findings highlight the dominant role of land-use change, particularly deforestation, in shaping runoff dynamics and flood risks in the Wardha River sub-basin, compared with the broader but less intense influence of rainfall variability.
- Open Access
- 20 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 203762-203779
Kokkidio: Fast, Expressive, Portable Code, Based on Kokkos and Eigen
Kokkidio is a newly developed C++ template library that bridges the performance portability framework Kokkos and its strength in utilising GPUs with the expressive syntax and CPU optimisations of the linear algebra library Eigen. The unified abstractions provided by Kokkidio enable both simple data management as well as clear, succinct compute code in kernel functors, where a novel iteration/functor parameter performs target-specific grouping of operations. When compiled for CPUs, this approach distributes contiguous data segments to OpenMP threads, while preserving Eigen’s loop abstractions to enable explicit vectorisation, thus utilising both thread- and data-level parallelism. Compiling the same source code for accelerators, such as GPUs, leverages Kokkos’ parallel dispatch instead, achieving compatibility with all prevalent GPU backends (CUDA, ROCm, SYCL, ...). A comprehensive evaluation across GPUs and CPUs by all major vendors shows Kokkidio providing significantly improved performance portability over Kokkos:With the fraction of best observed runtime as the efficiency metric, Kokkidio achieves a near-optimal harmonic mean of 95% efficiency across all microbenchmarks and tested hardware, offering a marked improvement over to Kokkos’ 59%, thus effectively merging Kokkos’ GPU and Eigen’s CPU performance, powerful math functions, and legibility into one framework.
- 20 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 203762-203779
Open AccessKokkidio: Fast, Expressive, Portable Code, Based on Kokkos and Eigen
Kokkidio is a newly developed C++ template library that bridges the performance portability framework Kokkos and its strength in utilising GPUs with the expressive syntax and CPU optimisations of the linear algebra library Eigen. The unified abstractions provided by Kokkidio enable both simple data management as well as clear, succinct compute code in kernel functors, where a novel iteration/functor parameter performs target-specific grouping of operations. When compiled for CPUs, this approach distributes contiguous data segments to OpenMP threads, while preserving Eigen’s loop abstractions to enable explicit vectorisation, thus utilising both thread- and data-level parallelism. Compiling the same source code for accelerators, such as GPUs, leverages Kokkos’ parallel dispatch instead, achieving compatibility with all prevalent GPU backends (CUDA, ROCm, SYCL, ...). A comprehensive evaluation across GPUs and CPUs by all major vendors shows Kokkidio providing significantly improved performance portability over Kokkos:With the fraction of best observed runtime as the efficiency metric, Kokkidio achieves a near-optimal harmonic mean of 95% efficiency across all microbenchmarks and tested hardware, offering a marked improvement over to Kokkos’ 59%, thus effectively merging Kokkos’ GPU and Eigen’s CPU performance, powerful math functions, and legibility into one framework.
- Open Access
- 3 March 2025
- Published by Springer Nature in Environmental Fluid Mechanics
- Vol. 25 (2) , 1-22
Three-dimensional modeling of diffusion-gravity flows in ice-covered lakes
When a solid inclined surface is submerged in a quiescent stratified fluid, the combined effects of buoyancy forces and diffusion generate an upward gravity flow along the slope. Thermally stratified ice-covered lakes remain in a nearly quiescent state and are potentially prone to this effect. We use three-dimensional hydrodynamic modeling to investigate the diffusion-gravity flow and its impact on lake-wide circulation in idealized ice-covered lakes. The qualitative characteristics of the boundary flow were adequately simulated by the model, supported by a good agreement with theoretical predictions. In enclosed lakes, the modeled diffusion-driven boundary flow generates residual circulation, which overturns the entire lake water column within 1 to 6 months, suggesting a significant contribution of this mechanism to heat and mass transport in lakes with long ice-covered seasons. When the insulation boundary condition is lifted and additional buoyancy is produced by heat flux from lake sediment, a counterflow emerges, resulting in a circulation pattern characterized by the superposition of two opposing boundary flows. At flux magnitudes exceeding one watt per square meter, the counterflow can entirely replace the diffusion-driven circulation. Due to the small magnitudes of these flows, the Coriolis effect substantially influences circulation, partially transforming radial flow into rotational lake-wide "gyres." The number and rotational direction of these gyres depend on the relative contribution of bottom heat flux. The results provide a framework for designing field studies in real lakes and investigating circulation effects on the transport of dissolved matter, such as nutrients, oxygen and greenhouse gases in ice-covered lakes.
- 3 March 2025
- Published by Springer Nature in Environmental Fluid Mechanics
- Vol. 25 (2) , 1-22
Open AccessThree-dimensional modeling of diffusion-gravity flows in ice-covered lakes
When a solid inclined surface is submerged in a quiescent stratified fluid, the combined effects of buoyancy forces and diffusion generate an upward gravity flow along the slope. Thermally stratified ice-covered lakes remain in a nearly quiescent state and are potentially prone to this effect. We use three-dimensional hydrodynamic modeling to investigate the diffusion-gravity flow and its impact on lake-wide circulation in idealized ice-covered lakes. The qualitative characteristics of the boundary flow were adequately simulated by the model, supported by a good agreement with theoretical predictions. In enclosed lakes, the modeled diffusion-driven boundary flow generates residual circulation, which overturns the entire lake water column within 1 to 6 months, suggesting a significant contribution of this mechanism to heat and mass transport in lakes with long ice-covered seasons. When the insulation boundary condition is lifted and additional buoyancy is produced by heat flux from lake sediment, a counterflow emerges, resulting in a circulation pattern characterized by the superposition of two opposing boundary flows. At flux magnitudes exceeding one watt per square meter, the counterflow can entirely replace the diffusion-driven circulation. Due to the small magnitudes of these flows, the Coriolis effect substantially influences circulation, partially transforming radial flow into rotational lake-wide "gyres." The number and rotational direction of these gyres depend on the relative contribution of bottom heat flux. The results provide a framework for designing field studies in real lakes and investigating circulation effects on the transport of dissolved matter, such as nutrients, oxygen and greenhouse gases in ice-covered lakes.