A spatiotemporal deep fusion model for merging satellite and gauge precipitation in China

Wu, Hongcai; Yang, Qinli; Liu, Jiaming; Wang, Guoqing

doi:10.1016/j.jhydrol.2020.124664

Cited by 141 publications

(68 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wehbe et al [51] used ANN to develop MSPD by merging weather radar, SPDs, and RGs. Wu et al [52] developed MSPD using SPDs and RGs by employing a spatiotemporal deep fusion (SDF) algorithm over China. Yin et al [53] developed MSPD using the cuckoo search (CS) algorithm and BMA at threestages over China.…”

Section: Introductionmentioning

confidence: 99%

Assessment of Merged Satellite Precipitation Datasets in Monitoring Meteorological Drought over Pakistan

2021

View full text Add to dashboard Cite

The current study evaluates the potential of merged satellite precipitation datasets (MSPDs) against rain gauges (RGs) and satellite precipitation datasets (SPDs) in monitoring meteorological drought over Pakistan during 2000–2015. MSPDs evaluated in the current study include Regional Weighted Average Least Square (RWALS), Weighted Average Least Square (WALS), Dynamic Clustered Bayesian model Averaging (DCBA), and Dynamic Bayesian Model Averaging (DBMA) algorithms, while the set of SPDs is Global Precipitation Measurement (GPM)-based Integrated Multi-Satellite Retrievals for GPM (IMERG-V06), Tropical Rainfall Measurement Mission (TRMM) Multi-Satellite Precipitation Analysis (TMPA 3B42 V7), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), and ERA-Interim (re-analyses dataset). Several standardized precipitation indices (SPIs), including SPI-1, SPI-3, and SPI-12, are used to evaluate the performances of RGs, SPDs, and MSPDs across Pakistan as well as on a regional scale. The Mann–Kendall (MK) test is used to assess the trend of meteorological drought across different climate regions of Pakistan using these SPI indices. Results revealed higher performance of MSPDs than SPDs when compared against RGs for SPI estimates. The seasonal evaluation of SPIs from RGs, MSPDs, and SPDs in a representative drought year (2008) revealed mildly to moderate wetness in monsoon season while mild to moderate drought in winter season across Pakistan. However, the drought severity ranges from mild to severe drought in different years across different climate regions. MAPD (mean absolute percentage difference) shows high accuracy (MAPD <10%) for RWALS-MSPD, good accuracy (10% < MAPD <20%) for WALS-MSPD and DCBA-MSPD, while good to reasonable accuracy (20% < MAPD < 50%) for DCBA in different climate regions. Furthermore, MSPDs show a consistent drought trend as compared with RGs, while SPDs show poor performance. Overall, this study demonstrated significantly improved performance of MSPDs in monitoring the meteorological drought.

show abstract

Section: Introductionmentioning

confidence: 99%

Assessment of Merged Satellite Precipitation Datasets in Monitoring Meteorological Drought over Pakistan

2021

View full text Add to dashboard Cite

show abstract

“…daily or hourly) is challenging and valuable (Chen et al, 2020b;R. Lima et al, 2021;Sun and Lan, 2021;Wu et al, 2020). Whether our method could be applied on these scales might need validation.…”

Section: Further Researchesmentioning

confidence: 99%

“…To alleviate the inherent biases, many calibration methods have been proposed for merging gauge observations and satellite-based precipitation to improve the accuracy and spatial coverage of precipitation, such as nonparametric kernel smoothing method (Li and Shao, 2010), geographical difference analysis (GDA) (Cheema and Bastiaanssen, 2012), geographical ratio analysis (GRA) (Duan and Bastiaanssen, 2013), conditional merging (CM) (Berndt et al, 2014), quantile mapping (Chen et al, 2013;Zhang and Tang, 2015), optimal interpolation (Lu et al, 2020;Wu et al, 2018;Xie and Xiong, 2011), GWR (Chao et al, 2018;Chen et al, 2018;Lu et al, 2019) and geostatistical interpolation (Park et al, 2017). However, these methods are based on some strict assumptions which might not be satisfied in practice (Wu et al, 2020;Zhang et al, 2021). Moreover, the precipitation-related environmental variables were not taken into account.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, the precipitation-related environmental variables were not taken into account. To this end, ML-based calibration methods have become popular, such as Quantile Regression Forests (QRF) (Bhuiyan et al, 2018), ANN (Pham et al, 2020;Yang and Luo, 2014), deep neural network (Tao et al, 2016), RF (Baez-Villanueva et al, 2020), convolutional neural network (CNN) (Wu et al, 2020), SVM and extreme learning machine (Zhang et al, 2021). In contrast, RF with excellent results has been widely adopted in plenty of studies (Baez-Villanueva et al, 2020;Bhuiyan et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Easy-to-use spatial Random Forest-based downscaling-calibration method for producing high resolution and accurate precipitation data

Chen

2021

Preprint

View full text Add to dashboard Cite

Abstract. High resolution and accurate precipitation data is significantly important for numerous hydrological applications. To enhance the spatial resolution and accuracy of satellite-based precipitation products, an easy-to-use downscaling-calibration method based on spatial Random Forest (SRF) is proposed in this paper, where the spatial autocorrelation between precipitation measurements is taken into account. The proposed method consists of two main stages. Firstly, the satellite-based precipitation was downscaled by SRF with the incorporation of some high-resolution covariates including latitude, longitude, DEM, NDVI, terrain slope, aspect, relief, and land surface temperatures. Then, the downscaled precipitation was calibrated by SRF with rain gauge observations and the aforementioned high-resolution variables. The monthly Integrated MultisatellitE Retrievals for Global Precipitation Measurement (IMERG) located in Sichuan province, China from 2015 to 2019 was processed using our method and its results were compared with those of some classical methods including geographically weighted regression (GWR), artificial neural network (ANN), random forest (RF), kriging interpolation only on gauge measurements, bilinear interpolation-based downscaling and then SRF-based calibration (Bi-SRF), and SRF-based downscaling and then geographical difference analysis (GDA)-based calibration (SRF-GDA). Results show that: (1) the proposed method outperforms the other methods as well as the original IMERG; (2) the monthly-based SRF estimation is slightly more accurate than the annual-based SRF fraction disaggregation method; (3) SRF-based downscaling and calibration preforms better than bilinear downscaling (Bi-SRF) and GDA-based calibration (SRF-GDA); (4) kriging seems more accurate than GWR and ANN in terms of quantitative accuracy measures, whereas its precipitation map cannot capture the detailed spatial precipitation patterns; and (5) among the predictors for calibration, the precipitation interpolated by kriging on the gauge measurements is the most important variable, indicating the significance for the inclusion of spatial autocorrelation information in gauge measurements.

show abstract

“…e spatial distributions of the metrics for the three models' 6-minute precipitation estimations over the study area are shown in Figure 7. Based on the observations from the 260 Automatic Weather Stations (AWSs), the CC, RMSE, MAE, and MB of the precipitation estimates obtained using the Z-R (1), MLP model, and FFNet-LSTM (Loss_Sum) were calculated for each gauge, and the gauge values were interpolated into others using the inverse distance weighting (IDW) interpolation to obtain the spatial distribution of the metrics [70]. As shown in Figure 8, the correlation coefficients of the Z-R (1) and MLP model have a large area of low values in the center of the study area, while the FFNet-LSTM (Loss_Sum) reduces the area of the low CC values in the center area and further improves the overall correlation coefficient of the study area.…”

Section: Spatial Distribution Of the Metrics For The Different Modelsmentioning

confidence: 99%

Offline Single-Polarization Radar Quantitative Precipitation Estimation Based on a Spatiotemporal Deep Fusion Model

Zhang

Chen

Tian

et al. 2021

Advances in Meteorology

View full text Add to dashboard Cite

Quantitative precipitation estimation (QPE) based on Doppler radar plays an important role in severe weather monitoring, industrial and agricultural production, and natural disaster prediction and prevention. However, the temporal and spatial variability of precipitation leads to large errors in radar estimates of mixed precipitation. To improve the accuracy of radar QPE, we propose an offline spatiotemporal deep fusion model that uses the reflectivity data of the Shijiazhuang Doppler radar Z9311 and the precipitation data from 17 national weather stations (NWSs) and 260 automatic weather stations (AWSs). Considering the abrupt spatial changes in precipitation, a three-dimensional radar data structure is proposed, and the spatial features of multielevation and multiscale radar data are extracted and merged using the feature fusion network (FFNet). Finally, the time dependence of the precipitation is captured using the long short-term memory (LSTM) network, and the precipitation estimation is obtained. Based on a comparison of the results of the proposed model (FFNet-LSTM) with those of the ordinary kriging (OK) interpolation, two Z-R relationship, the multilayer perceptron (MLP), the LSTM, and the FFNet, the proposed method is superior to these models, has a promising performance, and is a general-purpose rainfall algorithm.

show abstract

A spatiotemporal deep fusion model for merging satellite and gauge precipitation in China

Cited by 141 publications

References 53 publications

Assessment of Merged Satellite Precipitation Datasets in Monitoring Meteorological Drought over Pakistan

Assessment of Merged Satellite Precipitation Datasets in Monitoring Meteorological Drought over Pakistan

Easy-to-use spatial Random Forest-based downscaling-calibration method for producing high resolution and accurate precipitation data

Offline Single-Polarization Radar Quantitative Precipitation Estimation Based on a Spatiotemporal Deep Fusion Model

Contact Info

Product

Resources

About