Data fusion of satellite imagery and downscaling for generating highly fine-scale precipitation

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 631, P. 130665 - 130665

Published: Jan. 26, 2024

Language: Английский

---

A deep learning-based framework for multi-source precipitation fusion

Remote Sensing of Environment, Journal Year: 2023, Volume and Issue: 295, P. 113723 - 113723

Published: July 18, 2023

Language: Английский

---

An explainable two-stage machine learning approach for precipitation forecast

Journal of Hydrology, Journal Year: 2023, Volume and Issue: 627, P. 130375 - 130375

Published: Oct. 21, 2023

A common post-processing approach to improve precipitation forecasts is use machine learning models such as artificial neural networks (more specifically, multi-layer perceptrons) black-box systems. These utilize different sources of observations or predictors generate an improved forecast in terms desired metrics. However, most existing studies employ a single-stage regression model without considering explainability. The small number with two-stage that combine classification and binary still lack explainable intelligence. Therefore, this study proposes prediction system which (i) composed two stages for better predictions, (ii) compares the utility multi-class over regression, (iii) explainable, unlike prior studies, individual predictions learning-based are interpretable by humans. proposed first estimates intensity category using stage later utilizes information model, second stage, obtain daily magnitude. utilized made humanly (i.e., explainable) providing insight into model-wide importance generation processes (instance-level explanation). compared against approaches quality explainability, where station-based used ground truth datasets. Experiments show yields significant improvement (on average, RMSE reduced 10.50%, correlation between numerical observed values increased 7.5%) best-performing physical predictor (ECMWF). Analysis explainability provides insights decisions our approach, e.g., usefulness seasonality-related parameters, each task (regression classification).

Language: Английский

---

Unravelling and improving the potential of global discharge reanalysis dataset in streamflow estimation in ungauged basins

Journal of Cleaner Production, Journal Year: 2023, Volume and Issue: 419, P. 138282 - 138282

Published: July 27, 2023

Language: Английский

---

Downscaling and merging multiple satellite precipitation products and gauge observations using random forest with the incorporation of spatial autocorrelation

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 632, P. 130919 - 130919

Published: Feb. 15, 2024

Language: Английский

---

Precipitation data merging via machine learning: Revisiting conceptual and technical aspects

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 637, P. 131424 - 131424

Published: May 25, 2024

The development of accurate precipitation products with wide spatio-temporal coverage is crucial for a range applications. In this context, data merging (PDM) that entails the blending satellite-based estimates ground-based measurements holds prominent position, while currently there an increasing trend in deployment machine learning (ML) algorithms such endeavors. light recent advances field, work discusses key aspects PDM problem associated with: a) conceptual formulation problem, closely related to training ML models and their predictive capacity, b) selection fused, latency final product operational applicability method, c) efficiency single-step two-step approaches, former one treating via only regression latter combined use classification algorithms. By formulating as prediction we define assess two different strategies models, termed full per time step strategy, which entail building single or several respectively. Furthermore, performance allows predictions both spatial temporal dimensions, assessed context merging. each three scenarios, popular ensemble tree-based algorithms, i.e., random forest, gradient boosting extreme algorithm, are employed resulting nine merged products. To provide empirical evidence, employ datacube composed by daily observations, reanalysis estimates, well auxiliary covariates, from 1009 uniformly distributed cells (representative sampling area 25 × km), over four countries around world (Australia, USA, India Italy). large-scale experiment indicates that: (i) strategy competitive alternative since it enables methods improved accuracy, respect metrics reproduction statistics, but also higher capability applicability, (ii) much better occurrence characteristics, reflected improvement relevant categorical metrics, probability autocorrelation coefficient, (iii) no significant difference was noticed

Language: Английский

---

Machine Learning-Based Bias Correction of Precipitation Measurements at High Altitude

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(8), P. 2180 - 2180

Published: April 20, 2023

Accurate precipitation measurements are essential for understanding hydrological processes in high-altitude regions. Conventional gauge often yield large underestimations of actual precipitation, prompting the development statistical methods to correct measurement bias. However, complex conditions at high altitudes pose additional challenges methods. To improve correction areas, we selected Yakou station, situated an altitude 4147 m on Tibetan plateau, as study site. In this study, employed machine learning method XGBoost regression using meteorological variables and remote sensing data, including Global Satellite Mapping Precipitation (GSMaP), Integrated Multi-satellitE Retrievals GPM (IMERG) Climate Hazards Group InfraRed with Station data (CHIRPS). Additionally, examined transferability between different stations our site, Norway, United States. Our results show that station experiences a underestimation magnitude 51.4%. This is significantly higher than similar taken Arctic or lower altitudes. Furthermore, datasets underestimated when compared Double Fence Intercomparison Reference (DFIR) observation. findings suggest outperformed traditional accuracy metrics frequency distribution. Introducing especially GSMaP could potentially replace role situ wind speed correction, highlighting potential correcting rather Moreover, indicate demonstrated better cross-validated sites located countries. offers promising strategy obtaining more accurate

Language: Английский

---

Multi-source precipitation estimation using machine learning: Clarification and benchmarking

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 635, P. 131195 - 131195

Published: April 6, 2024

Language: Английский

---

Comparison of Tree-Based Ensemble Algorithms for Merging Satellite and Earth-Observed Precipitation Data at the Daily Time Scale

Hydrology, Journal Year: 2023, Volume and Issue: 10(2), P. 50 - 50

Published: Feb. 12, 2023

Merging satellite products and ground-based measurements is often required for obtaining precipitation datasets that simultaneously cover large regions with high density are more accurate than pure products. Machine statistical learning regression algorithms regularly utilized in this endeavor. At the same time, tree-based ensemble adopted various fields solving problems accuracy low computational costs. Still, information on which algorithm to select correcting contiguous United States (US) at daily time scale missing from literature. In study, we worked towards filling methodological gap by conducting an extensive comparison between three of category interest, specifically random forests, gradient boosting machines (gbm) extreme (XGBoost). We used data PERSIANN (Precipitation Estimation Remotely Sensed Information using Artificial Neural Networks) IMERG (Integrated Multi-satellitE Retrievals GPM) gridded datasets. also earth-observed Global Historical Climatology Network (GHCNd) database. The experiments referred entire US additionally included application linear benchmarking purposes. results suggest XGBoost best-performing among those compared. Indeed, mean relative improvements it provided respect (for case latter was run predictors as XGBoost) equal 52.66%, 56.26% 64.55% different predictor sets), while respective values 37.57%, 53.99% 54.39% 34.72%, 47.99% 62.61% gbm. Lastly, useful context investigated.

Language: Английский

---

Multisource Precipitation Data Merging Using a Dual-Layer ConvLSTM Model

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(3), P. 546 - 546

Published: Feb. 5, 2025

Precipitation is a key component of the water cycle. Different precipitation data sources have strengths and weaknesses. To combine these achieve accurate data, this study introduces dual-layer neural network (D-ConvLSTM) based on convolutional long short-term memory (ConvLSTM) that integrates ground station (1 h interval) grid generated by China Meteorological Administration Multi-source merged Analysis System (CMPAS, 1 interval, 0.05° × 0.05°) through two-layer for identification correction. evaluate performance proposed model, D-ConvLSTM, optimal interpolation (OI), single-layer ConvLSTM model are evaluated in Dadu River Basin, China. The results show D-ConvLSTM outperforms CMPAS all metrics compared with OI ConvLSTM, improvements 18.9% 19.8% critical success index (CSI) Kling–Gupta efficiency (KGE), respectively. enhances gridded under various conditions, including areas without different intensities, regions. Furthermore, analyzes impact training distribution adjusting distribution. analysis reveals ratio dry to wet set affects model’s performance. overestimation underestimation observations influences value This offers new merging provides insights enhancing accuracy merging.

Language: Английский

---