Deep Learning for Precipitation Estimation from Satellite and Rain Gauges Measurements

Arthur Moraux,Steven Dewitte,Bruno Cornelis,Adrian Munteanu

doi:10.3390/rs11212463

Arthur Moraux, Steven Dewitte + Show 2 more

Open Access

https://doi.org/10.3390/rs11212463

Copy DOI

Abstract

This paper proposes a multimodal and multi-task deep-learning model for instantaneous precipitation rate estimation. Using both thermal infrared satellite radiometer and automatic rain gauge measurements as input, our encoder–decoder convolutional neural network performs a multiscale analysis of these two modalities to estimate simultaneously the rainfall probability and the precipitation rate value. Precipitating pixels are detected with a Probability Of Detection (POD) of 0.75 and a False Alarm Ratio (FAR) of 0.3. Instantaneous precipitation rate is estimated with a Root Mean Squared Error (RMSE) of 1.6 mm/h.

Highlights

Instantaneous precipitation rate estimation is an important problem for meteorological, climatological and hydrological applications
We evaluated the ability of our model to detect precipitation and to accurately estimate their rate by computing various scores already used in previous studies [9,10]
The considered classification scores include the Probability Of Detection (POD), the False Alarm Ratio (FAR), the Probability of False Detection (POFD), the Accuracy (ACC), the Critical Success Index (CSI), the Gilbert Skill Score (GSS), the Heidke Skill Score (HSS), the Hanssen–Kuipers Discriminant (HKD) and the F1 score (F1). All these classification scores (Table 4) are computed from the contingency table made from the observations of the rain gauges and the estimations of our model (Table 5)

Summary

Introduction

Instantaneous precipitation rate estimation is an important problem for meteorological, climatological and hydrological applications. It forms the basis for short-term precipitation forecasting, called nowcasting [1]. Rain gauges are considered to be the reference devices for the measurement of the amount of precipitation at ground level. Climatological rain gauges are simple recipients, manually read out once per day. There are networks of automatic stations, able to report rainfall quantity every 5 or 10 min. The main drawback of rain gauges is their lack of spatial representativity, being only point measurements

Methods

Results

Discussion

Conclusion