Merging Satellite and Gauge-Measured Precipitation Using LightGBM With an Emphasis on Extreme Quantiles DOI Creative Commons
Hristos Tyralis, Georgia Papacharalampous, Nikolaos Doulamis

et al.

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Journal Year: 2023, Volume and Issue: 16, P. 6969 - 6979

Published: Jan. 1, 2023

Knowing the actual precipitation in space and time is critical hydrological modelling applications, yet spatial coverage with rain gauge stations limited due to economic constraints. Gridded satellite datasets offer an alternative option for estimating by covering uniformly large areas, albeit related estimates are not accurate. To improve estimates, machine learning applied merge gauge-based measurements gridded products. In this context, observed plays role of dependent variable, while data play predictor variables. Random forests dominant algorithm relevant applications. those predictions settings, point (mostly mean or median conditional distribution) variable issued. The aim manuscript solve problem probabilistic prediction emphasis on extreme quantiles interpolation settings. Here we propose, issuing using Light Gradient Boosting Machine (LightGBM). LightGBM a boosting algorithm, highlighted prize-winning entries forecasting competitions. assess LightGBM, contribute large-scale application that includes merging daily contiguous US PERSIANN GPM-IMERG data. We focus probability distribution where outperforms quantile regression (QRF, variant random forests) terms score at quantiles. Our study offers understanding settings learning.

Language: Английский

Improving near-real-time satellite precipitation products through multistage modified schemes DOI
Chengcheng Meng, Xingguo Mo, Suxia Liu

et al.

Atmospheric Research, Journal Year: 2023, Volume and Issue: 292, P. 106875 - 106875

Published: June 15, 2023

Language: Английский

Citations

9
Machine learning approaches for reconstructing gridded precipitation based on multiple source products DOI Creative Commons
Giang V. Nguyen, Xuan-Hien Le, Linh Nguyen Van

et al.

Journal of Hydrology Regional Studies, Journal Year: 2023, Volume and Issue: 48, P. 101475 - 101475

Published: July 14, 2023

South Korea is situated in the northeastern region of Asia Recent technological developments have enabled multi-source precipitation products (MSPs), including satellite-based and model-based, to be useful data sources for quantifying spatiotemporal variations precipitation. Unfortunately, main limitation MSPs potential applications inheritance errors with high uncertainty. To tackle this problem, capabilities six machine learning algorithms (Ridge Linear Regression, k-Nearest Neighbors, Support Vector Gradient Boosting Decision Tree, Light Machine, Random Forest) produce new product by merging ground-based investigated. Ground-based from 2003 2017 were utilized train valid process. The robustness ML was highlighted using several evaluation metrics such as continuous indices (modified Kling-Gupta Efficiency root mean square error) categorical (probability detection, false alarm rate, critical success index). results indicate that (1) approaches can merge observed accurately estimate rainfall, particularly basins sparsely distributed rain gauge stations. (2) merged generated showed significant agreement accuracy observation considering rainfall intensity estimation improved capability detecting non-rain events over Korea.

Language: Английский

Citations

9
A D-vine copula-based quantile regression towards merging satellite precipitation products over rugged topography: a case study in the upper Tekeze–Atbara Basin DOI Creative Commons
Mohammed Abdallah, Ke Zhang, Lijun Chao

et al.

Hydrology and earth system sciences, Journal Year: 2024, Volume and Issue: 28(5), P. 1147 - 1172

Published: March 7, 2024

Abstract. Precipitation is a vital key element in various studies of hydrology, flood prediction, drought monitoring, and water resource management. The main challenge conducting over remote regions with rugged topography that weather stations are usually scarce unevenly distributed. However, open-source satellite-based precipitation products (SPPs) suitable resolution provide alternative options these data-scarce regions, which typically associated high uncertainty. To reduce the uncertainty individual satellite products, we have proposed D-vine copula-based quantile regression (DVQR) model to merge multiple SPPs rain gauges (RGs). DVQR was employed during 2001–2017 summer monsoon seasons compared two other methods based on multivariate linear (MLQR) Bayesian averaging (BMAQ) techniques, respectively, traditional merging – simple modeling average (SMA) one-outlier-removed (OORA) using descriptive categorical statistics. Four been considered this study, namely, Tropical Applications Meteorology SATellite (TAMSAT v3.1), Climate Prediction Center MORPHing Product Data Record (CMORPH-CDR), Global Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG v06), Estimation from Remotely Sensed Information Artificial Neural Networks (PERSIANN-CDR). bilinear (BIL) interpolation technique applied downscale coarse fine spatial (1 km). rugged-topography region upper Tekeze–Atbara Basin (UTAB) Ethiopia selected as study area. results indicate data estimates DVQR, MLQR, BMAQ models outperform downscaled SPPs. Monthly evaluations reveal all perform better July September than June August due variability. exhibit higher accuracy UTAB. substantially improved statistical metrics (CC = 0.80, NSE 0.615, KGE 0.785, MAE 1.97 mm d−1, RMSE 2.86 PBIAS 0.96 %) MLQR models. did not respect probability detection (POD) false-alarm ratio (FAR), although it had best frequency bias index (FBI) critical success (CSI) among Overall, newly approach improves quality demonstrates value such

Language: Английский

Citations

2
Added value of merging techniques in precipitation estimates relative to gauge-interpolation algorithms of varying complexity DOI
Yanqiu Hu, Ling Zhang

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 645, P. 132214 - 132214

Published: Oct. 22, 2024

Language: Английский

Citations

2
Merging Satellite and Gauge-Measured Precipitation Using LightGBM With an Emphasis on Extreme Quantiles DOI Creative Commons
Hristos Tyralis, Georgia Papacharalampous, Nikolaos Doulamis

et al.

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Journal Year: 2023, Volume and Issue: 16, P. 6969 - 6979

Published: Jan. 1, 2023

Knowing the actual precipitation in space and time is critical hydrological modelling applications, yet spatial coverage with rain gauge stations limited due to economic constraints. Gridded satellite datasets offer an alternative option for estimating by covering uniformly large areas, albeit related estimates are not accurate. To improve estimates, machine learning applied merge gauge-based measurements gridded products. In this context, observed plays role of dependent variable, while data play predictor variables. Random forests dominant algorithm relevant applications. those predictions settings, point (mostly mean or median conditional distribution) variable issued. The aim manuscript solve problem probabilistic prediction emphasis on extreme quantiles interpolation settings. Here we propose, issuing using Light Gradient Boosting Machine (LightGBM). LightGBM a boosting algorithm, highlighted prize-winning entries forecasting competitions. assess LightGBM, contribute large-scale application that includes merging daily contiguous US PERSIANN GPM-IMERG data. We focus probability distribution where outperforms quantile regression (QRF, variant random forests) terms score at quantiles. Our study offers understanding settings learning.

Language: Английский

Citations

6