Precipitation Estimates from MSG SEVIRI Daytime, Nighttime, and Twilight Data with Random Forests

Kühnlein, Meike; Appelhans, Tim; Thies, Boris; Nauß, Thomas

doi:10.1175/jamc-d-14-0082.1

Cited by 73 publications

(87 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The modelling methodology follows the study of Kühnlein et al (2014a, b) who used the spectral channels of MSG SE-VIRI to train a random forest model that is able to spatially estimate rainfall areas and rainfall rates over Germany. Based on this study, Meyer et al (2016) have shown that neural networks outperform the initially used random forest algorithm.…”

Section: Model Strategies For Rainfall Estimation 231 General Modelmentioning

confidence: 99%

“…Kühnlein et al (2014a, b) and Meyer et al (2016) presented a methodology to estimate rainfall from optical Meteosat Second Generation (MSG) Spinning Enhanced Visible and InfraRed Imager (SEVIRI) data for Germany. In this approach, machine learning algorithms were used to relate the spectral properties of MSG to reliable radar data as a ground truth.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Satellite-based high-resolution mapping of rainfall over southern Africa

2017

View full text Add to dashboard Cite

Abstract. A spatially explicit mapping of rainfall is necessary for southern Africa for eco-climatological studies or nowcasting but accurate estimates are still a challenging task. This study presents a method to estimate hourly rainfall based on data from the Meteosat Second Generation (MSG) Spinning Enhanced Visible and Infrared Imager (SE-VIRI). Rainfall measurements from about 350 weather stations from 2010-2014 served as ground truth for calibration and validation. SEVIRI and weather station data were used to train neural networks that allowed the estimation of rainfall area and rainfall quantities over all times of the day. The results revealed that 60 % of recorded rainfall events were correctly classified by the model (probability of detection, POD). However, the false alarm ratio (FAR) was high (0.80), leading to a Heidke skill score (HSS) of 0.18. Estimated hourly rainfall quantities were estimated with an average hourly correlation of ρ = 0.33 and a root mean square error (RMSE) of 0.72. The correlation increased with temporal aggregation to 0.52 (daily), 0.67 (weekly) and 0.71 (monthly). The main weakness was the overestimation of rainfall events. The model results were compared to the Integrated Multi-satellitE Retrievals for GPM (IMERG) of the Global Precipitation Measurement (GPM) mission. Despite being a comparably simple approach, the presented MSGbased rainfall retrieval outperformed GPM IMERG in terms of rainfall area detection: GPM IMERG had a considerably lower POD. The HSS was not significantly different compared to the MSG-based retrieval due to a lower FAR of GPM IMERG. There were no further significant differences between the MSG-based retrieval and GPM IMERG in terms of correlation with the observed rainfall quantities. The MSG-based retrieval, however, provides rainfall in a higher spatial resolution. Though estimating rainfall from satellite data remains challenging, especially at high temporal resolutions, this study showed promising results towards improved spatio-temporal estimates of rainfall over southern Africa.

show abstract

Section: Model Strategies For Rainfall Estimation 231 General Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Satellite-based high-resolution mapping of rainfall over southern Africa

2017

View full text Add to dashboard Cite

show abstract

“…In fact, it is widely known that correlations between predictions and observations tend to improve with increasing spatial (e.g., [38]) and temporal aggregation (e.g., [36,40,47,54]). We nevertheless decided to focus solely on pixel-based instantaneous rain rates here because they are especially valuable for real-time nowcasting applications, as they can e.g., be used to fill in gaps during radar downtime and cover regions with poor or no radar coverage.…”

Section: Discussionmentioning

confidence: 99%

“…Tapiador et al [52] provide a general overview on satellite-based rainfall estimations by means of neural networks including a discussion about advantages as well as drawbacks of employing such a statistical-learning approach. In addition, the potential of other machine learning techniques such as random forests has been revealed for precipitation detection based on LEO MW [53] as well as for rainfall retrievals based on GEO IR and visible (VIS) channel input data [54]. Meyer et al [40] compare several machine learning algorithms (random forests, artificial neural networks, and support vector machines) and show that the choice of machine learning algorithm only marginally affects the skill of the rainfall retrieval.…”

Section: Introductionmentioning

confidence: 99%

Satellite-Based Rainfall Retrieval: From Generalized Linear Models to Artificial Neural Networks

et al. 2018

View full text Add to dashboard Cite

Abstract:In this study, we develop and compare satellite rainfall retrievals based on generalized linear models and artificial neural networks. Both approaches are used in classification mode in a first step to identify the precipitating areas (precipitation detection) and in regression mode in a second step to estimate the rainfall intensity at the ground (rain rate). The input predictors are geostationary satellite infrared (IR) brightness temperatures and Satellite Application Facility (SAF) nowcasting products which consist of cloud properties, such as cloud top height and cloud type. Additionally, a set of auxiliary location-describing input variables is employed. The output predictand is the ground-based instantaneous rain rate provided by the European-scale radar composite OPERA, that was additionally quality-controlled. We compare our results to a precipitation product which uses a single infrared (IR) channel for the rainfall retrieval. Specifically, we choose the operational PR-OBS-3 hydrology SAF product as a representative example for this type of approach.With generalized linear models, we show that we are able to substantially improve in terms of hits by considering more IR channels and cloud property predictors. Furthermore, we demonstrate the added value of using artificial neural networks to further improve prediction skill by additionally reducing false alarms. In the rain rate estimation, the indirect relationship between surface rain rates and the cloud properties measurable with geostationary satellites limit the skill of all models, which leads to smooth predictions close to the mean rainfall intensity. Probability matching is explored as a tool to recover higher order statistics to obtain a more realistic rain rate distribution.

show abstract

“…The unbalanced categories in the dependent variable can cause the prediction error between categories also to be unbalanced in random forests modeling [35,47]. Therefore, stratified sampling was used to down-sample the majority category, which has been demonstrated to work better than over-sampling the minority category [48].…”

Section: Samplingmentioning

confidence: 99%

Identification of Factors Influencing Locations of Tree Cover Loss and Gain and Their Spatio-Temporally-Variant Importance in the Li River Basin, China

Zhang

Qin

2016

Remote Sensing

View full text Add to dashboard Cite

Abstract:Intensive tree cover loss and gain have been significantly influencing the environment and society. It is essential to identify the potential factors and to evaluate their importance. A large body of literature has investigated the factors influencing tree cover loss, usually at the global or regional scale and focusing on the quantity issue: how are the rate and extent of tree cover loss influenced by different factors? This paper has two objectives. The first is to pinpoint factors influencing the locations of tree cover loss and gain (the location issue) at the pixel level. The second is to evaluate the heterogeneous importance of factors in two periods of 1991 through 2002 and 2002 through 2013 and in four counties within the Li River Basin, Guangxi Zhuang Autonomous Region, China. The random forests technique was adopted to model the responses of tree cover loss and gain probabilities of sampled pixels to initial landscape pattern factors, biophysical factors and proximity factors. A ranking of factor importance and a set of important factors were derived for each county and time period. The partial dependence plots were generated for the most important factors to reveal how exactly tree cover loss and gain probabilities change as influenced by these factors. The results confirmed that factor importance varied across time and space, and the variability of proximity factors and initial landscape pattern factors were more pronounced. The furthered understanding of the heterogeneous importance of different factors on the locations of tree cover loss and gain can support more sustainable forest management practices and the development of more effective policies regarding ecosystem conservation and economic development.

show abstract

Precipitation Estimates from MSG SEVIRI Daytime, Nighttime, and Twilight Data with Random Forests

Cited by 73 publications

References 60 publications

Satellite-based high-resolution mapping of rainfall over southern Africa

Satellite-based high-resolution mapping of rainfall over southern Africa

Satellite-Based Rainfall Retrieval: From Generalized Linear Models to Artificial Neural Networks

Identification of Factors Influencing Locations of Tree Cover Loss and Gain and Their Spatio-Temporally-Variant Importance in the Li River Basin, China

Contact Info

Product

Resources

About