Deep Learning for Precipitation Estimation from Satellite and Rain Gauges Measurements

2020 ◽  
Author(s):  
Arthur Moraux ◽  
Steven Dewitte ◽  
Bruno Cornelis ◽  
Adrian Munteanu
Keyword(s):  
Deep Learning ◽  
Rain Gauge ◽  
High Expectations ◽  
Rain Gauges ◽  
Weather And Climate ◽  
Learning Technique ◽  
Precipitation Estimation ◽  
The Many ◽  
Weather Observation

<p><span>In the coming years, Artificial Intelligence (AI), for which Deep Learning (DL) is an essential component, is expected to transform society in a way that is compared to the introduction of electricity or the introduction of the internet. The high expectations are founded on the many impressive results of recent DL studies for AI tasks (e.g. computer vision, text translation, image or text generation...). Also for weather and climate observations, a large potential for </span><span>AI</span><span> application exists. </span></p><p><span>We present the results of the recent paper [Moraux et al, 2019], which is one of the first demonstrations of the application </span><span>of </span><span>cutting edge deep learning technique</span><span>s</span><span> to a practical weather observation problem. We developed a multiscale encoder-decoder convolutional neural network using the three most relevant SEVIRI/MSG spectral images at 8.7, 10.8 and 12.0 micron and in situ rain gauge measurements as input. The network is trained to reproduce precipitation measured by rain gauges in Belgium, the Netherlands and Germany. Precipitating pixels are detected with a POD of 0.75 and a FAR of 0.3. Instantaneous precipitation rate is estimated with a RMSE of 1.6 mm/h.</span></p><p> </p><p><span>Reference:</span></p><p><span>[Moraux et al, 2019] Moraux, A.; Dewitte, S.; Cornelis, B.; Munteanu, A. Deep Learning for Precipitation Estimation from Satellite and Rain Gauges Measurements. </span><em><span>Remote Sens.</span></em> <span><strong>2019</strong></span><span>, </span><em><span>11</span></em><span>, 2463. </span></p>

scholarly journals Deep Learning for Precipitation Estimation from Satellite and Rain Gauges Measurements

2019 ◽  
Vol 11 (21) ◽  
pp. 2463
Author(s):  
Arthur Moraux ◽  
Steven Dewitte ◽  
Bruno Cornelis ◽  
Adrian Munteanu
Keyword(s):  
Deep Learning ◽  
Multiscale Analysis ◽  
Mean Squared Error ◽  
Rain Gauge ◽  
Probability Of Detection ◽  
Precipitation Rate ◽  
Rain Gauges ◽  
Squared Error ◽  
Precipitation Estimation ◽  
Deep Learning Model

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.

scholarly journals Impact of the Altitudinal Gradients of Precipitation on the Radar QPE Bias in the French Alps

Atmosphere ◽  
2019 ◽  
Vol 10 (6) ◽  
pp. 306 ◽  
Author(s):  
Dominique Faure ◽  
Guy Delrieu ◽  
Nicolas Gaussiat
Keyword(s):  
Ground Level ◽  
Radar Data ◽  
Rain Gauge ◽  
Altitudinal Gradients ◽  
French Alps ◽  
Rain Gauges ◽  
Quantitative Precipitation Estimation ◽  
Precipitation Estimation ◽  
Radar Data Processing

In the French Alps the quality of the radar Quantitative Precipitation Estimation (QPE) is limited by the topography and the vertical structure of precipitation. A previous study realized in all the French Alps, has shown a general bias between values of the national radar QPE composite and the rain gauge measurements: a radar QPE over-estimation at low altitude (+20% at 200 m a.s.l.), and an increasing underestimation at high altitudes (until −40% at 2100 m a.s.l.). This trend has been linked to altitudinal gradients of precipitation observed at ground level. This paper analyzes relative altitudinal gradients of precipitation estimated with rain gauges measurements in 2016 for three massifs around Grenoble, and for different temporal accumulations (yearly, seasonal, monthly, daily). Comparisons of radar and rain gauge accumulations confirm the bias previously observed. The parts of the current radar data processing affecting the bias value are pointed out. The analysis shows a coherency between the relative gradient values estimated at the different temporal accumulations. Vertical profiles of precipitation detected by a research radar installed at the bottom of the valley also show how the wide horizontal variability of precipitation inside the valley can affect the gradient estimation.

Validation of PERSIANN Precipitation Product Using TAWN Rain Gauge Network Over Different Agro-climatic Zones in Tamil Nadu

2021 ◽  
Vol 108 (september) ◽  
pp. 1-6
Author(s):  
Venkadesh Samykannu ◽  
◽  
Pazhanivelan S ◽  
Keyword(s):  
High Altitude ◽  
Tamil Nadu ◽  
Rain Gauge ◽  
Northeast Monsoon ◽  
Data Set ◽  
Climatic Zones ◽  
Rain Gauges ◽  
Precipitation Estimation ◽  
Cauvery Delta ◽  
Rain Gauge Network

Currently, several satellite-precipitation products were developed using multiple algorithms to estimate rainfall. This study carried out using Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) product over seven agro-climatic zones of Tamil Nadu during the northeast monsoon (NEM) season of October to December for 2015-2017 (three years) against 118 rain-gauges data of Tamil Nadu Agricultural Weather Network (TAWN). The performance compares aggregated seasonal scale of rainfall using continuous (CC, RMSE, and NRMSE) statistical approaches. It was observed that PERSIANN is accurate in the high-altitude hilly zone and the Cauvery delta zone. For 2015, 2016, and 2017, the correlation values were 0.77, 0.52, and 0.71, respectively. The highest RMSE value was measured for northeast zone (NEZ) during 2015 (222.17 mm), and the lowest was determined for 22.63 in the High-altitude hilly zone (HAHZ) during 2016 and NRMSE had less errors during all three seasons. The study concluded that the PERSIANN data set could be useful substitute for rain-gauge precipitation data.

Satellite Data Application to Cover Lack of In-situ Observations for Mapping Precipitation and Direct Runoff in Semi-arid Basin

2020 ◽  
Author(s):  
Mahdi Akbari ◽  
Ali Torabi Haghighi
Keyword(s):  
Hydrological Modeling ◽  
Rain Gauge ◽  
Model Parameters ◽  
Lake Basin ◽  
Lake Urmia ◽  
Direct Runoff ◽  
Rain Gauges ◽  
High Elevations ◽  
Semi Arid

<div> <p>Hydrological modeling in arid basins located in developing countries often lacks sufficient hydrological data because, e.g., rain gauges are typically absent at high elevations and inflow to ungauged areas around large closed lakes such as Lake Urmia is difficult to estimate. We tried to improve precipitation and runoff estimation in Lake Urmia, Iran as an arid basin using satellite-based data. We estimated precipitation using interpolation of rain gauge data by kriging, downscaling Tropical Rainfall Measuring Mission (TRMM), and cokriging interpolation of in-situ records with Remote Sensing (RS)-based data. Using RS-based data in estimations gave more precise results, by compensating for lack of data at high elevations. Cokriging interpolation of rain gauges by TRMM and Digitized Elevation Model (DEM) gave 4–9 mm lower Root Mean Square Error (RMSE) in different years compared with kriging. Downscaling TRMM improved its accuracy by 14 mm. Using the most accurate precipitation model, we modeled annual direct runoff with Kennessey and Soil Conservation Service Curve Number (SCS-CN) models. These models use land use, permeability, slope maps and climatic parameter (Ia) to represent the annual climatic condition of modeled basin in sense of wetness or dryness. In runoff modeling, Kennessey gave higher accuracy in annual scale. It was found that classification of years to wet, dry and normal states in Kennessey by default assumptions on Ia is not accurate enough for semi-arid basins so by solving this issue and calibration Kennessey model parameters, we made this model applicable for Urmia Lake basin. Calibrating Kennessey reduced the Normalized RMSE (NRMSE) from 1 in the standard model to 0.44. Direct runoff coefficient map by 1 km spatial resolution was generated by calibrated Kennessey. Validation by the closest gauges to the lake gave a NRMSE of 0.41 which approved the accuracy of modeling.</p> </div>

scholarly journals On precipitation measurements collected by a weather radar and a rain gauge network

2013 ◽  
Vol 13 (3) ◽  
pp. 605-623 ◽  
Author(s):  
S. Sebastianelli ◽  
F. Russo ◽  
F. Napolitano ◽  
L. Baldini
Keyword(s):  
Rain Gauge ◽  
Rainfall Amount ◽  
Rainfall Time Series ◽  
Adjustment Factor ◽  
Radar Rainfall ◽  
Rain Gauges ◽  
R Ratio ◽  
Precipitation Estimation ◽  
Layer Effect ◽  
One Year

Abstract. Many phenomena (such as attenuation and range degradation) can influence the accuracy of rainfall radar estimates. They introduce errors that increase as the distance from radar increases, thereby decreasing the reliability of radar estimates for applications that require quantitative precipitation estimation. The present paper evaluates radar error as a function of the range, in order to correct the rainfall radar estimates. The radar is calibrated utilizing data from the rain gauges. Then, the G/R ratio between the yearly rainfall amount measured in each rain gauge position during 2008 and the corresponding radar rainfall amount is calculated against the slant range. The trend of the G/R ratio shows two behaviours: a concave part due to the melting layer effect close to the radar location and an almost linear, increasing trend at greater distances. A best fitting line is used to find an adjustment factor, which estimates the radar error at a given range. The effectiveness of the methodology is verified by comparing pairs of rainfall time series that are observed simultaneously by collocated rain gauges and radar. Furthermore, the variability of the adjustment factor is investigated at the scale of event, both for convective and stratiform events. The main result is that there is not a univocal range error pattern, as it also depends on the characteristics of the considered event. On the other hand, the adjustment factor tends to stabilize itself for time aggregations of the order of one year or greater.

scholarly journals Advancing Precipitation Estimation and Streamflow Simulations in Complex Terrain with X-Band Dual-Polarization Radar Observations

2018 ◽  
Vol 10 (8) ◽  
pp. 1258 ◽  
Author(s):  
Marios Anagnostou ◽  
Efthymios Nikolopoulos ◽  
John Kalogiros ◽  
Emmanouil Anagnostou ◽  
Francesco Marra ◽  
...  
Keyword(s):  
Complex Terrain ◽  
Rain Gauge ◽  
Storm Events ◽  
Error Statistics ◽  
Radar Rainfall ◽  
Radar Observations ◽  
X Band ◽  
Precipitation Estimation ◽  
Rainfall Estimates

In mountain basins, the use of long-range operational weather radars is often associated with poor quantitative precipitation estimation due to a number of challenges posed by the complexity of terrain. As a result, the applicability of radar-based precipitation estimates for hydrological studies is often limited over areas that are in close proximity to the radar. This study evaluates the advantages of using X-band polarimetric (XPOL) radar as a means to fill the coverage gaps and improve complex terrain precipitation estimation and associated hydrological applications based on a field experiment conducted in an area of Northeast Italian Alps characterized by large elevation differences. The corresponding rainfall estimates from two operational C-band weather radar observations are compared to the XPOL rainfall estimates for a near-range (10–35 km) mountainous basin (64 km2). In situ rainfall observations from a dense rain gauge network and two disdrometers (a 2D-video and a Parsivel) are used for ground validation of the radar-rainfall estimates. Ten storm events over a period of two years are used to explore the differences between the locally deployed XPOL vs. longer-range operational radar-rainfall error statistics. Hourly aggregate rainfall estimates by XPOL, corrected for rain-path attenuation and vertical reflectivity profile, exhibited correlations between 0.70 and 0.99 against reference rainfall data and 21% mean relative error for rainfall rates above 0.2 mm h−1. The corresponding metrics from the operational radar-network rainfall products gave a strong underestimation (50–70%) and lower correlations (0.48–0.81). For the two highest flow-peak events, a hydrological model (Kinematic Local Excess Model) was forced with the different radar-rainfall estimations and in situ rain gauge precipitation data at hourly resolution, exhibiting close agreement between the XPOL and gauge-based driven runoff simulations, while the simulations obtained by the operational radar rainfall products resulted in a greatly underestimated runoff response.

scholarly journals Comparative analysis of TMPA and IMERG precipitation datasets in the arid environment of El- Qaa Plain, Sinai

2021 ◽  
Author(s):  
Mona Morsy ◽  
Thomas Scholten ◽  
Silas Michaelides ◽  
Erik Borg ◽  
Youssef Sherief ◽  
...  
Keyword(s):  
Light Intensity ◽  
Arid Environment ◽  
Rain Gauge ◽  
Data Set ◽  
Light Rain ◽  
Rain Gauges ◽  
Data Source ◽  
Local Water ◽  
Rain Events

<p>The replenishment of aquifers depends mainly on precipitation rates, which is of vital 19 importance for determining water budgets in arid and semi-arid regions. El-Qaa Plain in Sinai 20 Peninsula is such a region which experiences a constant population growth. The local water budget 21 equilibrium is negatively affected by relatively frequent light rain events. This study compares the 22 performance of two sets of satellite-based data of precipitation and in situ rainfall measurements. The 23 dates selected refer to rainfall events between 2015 and 2018. For this purpose, 0.1° and 0.25° spatial 24 resolution TMPA (TRMM Multi-satellite Precipitation Analysis) and IMERG (Integrated Multi-25 satellitE Retrievals for GPM) data were retrieved and analyzed, employing appropriate statistical 26 metrics. The best-performing data set was determined as the data source capable to most accurately 27 bridge gaps in the limited rain gauge records, embracing both frequent light-intensity rain events 28 and rarer heavy-intensity events. With light-intensity events the corresponding satellite-based data 29 sets differ the least and correlate more, while the greatest differences and weakest correlations are 30 noted for the heavy-intensity events. The satellite-based records best match those of the rain gauges 31 during light-intensity events, when compared to the heaviest ones. IMERG data exhibit a superior 32 performance than TMPA, in all rainfall intensities.</p>

scholarly journals Analysis of the Kernel Bandwidth Influence in the Double Smoothing Merging Algorithm to Improve Rainfall Fields in Poorly Gauged Basins

10.29007/2xp6 ◽  
2018 ◽  
Author(s):  
Nicolás Duque-Gardeazábal ◽  
David Zamora ◽  
Erasmo Rodríguez
Keyword(s):  
Climate Models ◽  
Water Cycle ◽  
Rain Gauge ◽  
Reanalysis Data ◽  
Sources Of Information ◽  
Rain Gauges ◽  
Merging Algorithm ◽  
Double Smoothing ◽  
Gauged Basins

Accurate estimates of precipitation are needed for many applications in hydrology as rainfall is one of the most influential variables of the water cycle. The common sources of information used to estimate rainfall fields are in situ rain gauges, remote sensing information and outputs from climate models. However, each of the above- mentioned sources has its own limitations, which can be reduced by blending information from these sources, in a product that takes advantage of the strengths of each dataset. In this research we study the double smoothing merging algorithm, creating a rainfall distributed product that combines remote sensed and reanalysis data, and information from a rain gauge network. The main objective of the study is to investigate the implications of varying the rain gauge density and configuration, on the merging parameters and global performance of the blended product. The results of a daily 3-year period experiment show that, although the errors in cross validation (CV) and against an independent dataset (IV) are in general low, the performance of the blended product and also the sensitivity of the parameters are highly influenced by the rain gauge configuration and density. The bandwidth merging parameters increase as the network density is artificially reduced.

scholarly journals Exploration of an adaptive merging scheme for optimal precipitation estimation over ungauged urban catchment

2016 ◽  
Vol 19 (2) ◽  
pp. 225-237 ◽  
Author(s):  
Sherien Fadhel ◽  
Miguel Angel Rico-Ramirez ◽  
Dawei Han
Keyword(s):  
Urban Areas ◽  
Radar Data ◽  
Rain Gauge ◽  
Optimum Number ◽  
Rainfall Estimation ◽  
Urban Catchment ◽  
Rain Gauges ◽  
Daily Data ◽  
Precipitation Estimation ◽  
Urban Catchments

Merging rain gauge and radar data improves the accuracy of precipitation estimation for urban areas. Since the rain gauge network around the ungauged urban catchment is fixed, the relevant question relates to the optimal merging area that produces the best rainfall estimation inside the catchment. Thus, an incremental radar-gauge merging was performed by gradually increasing the distance from the centre of the study area, the number of merging gauges around it and the radar domain. The proposed adaptive merging scheme is applied to a small urban catchment in west Yorkshire, Northern England, for 118 extreme events from 2007 to 2009. The performance of the scheme is assessed using four experimental rain gauges installed inside the study area. The result shows that there is indeed an optimum radar-gauge merging area and consequently there is an optimum number of rain gauges that produce the best merged rainfall data inside the study area. Different merging methods produce different results for both classified and unclassified rainfall types. Although the scheme was applied on daily data, it is applicable to other temporal resolutions. This study has importance for other studies such as urban flooding analysis, since it provides improved rainfall estimation for ungauged urban catchments.

scholarly journals A Deep Learning Multimodal Method for Precipitation Estimation

2021 ◽  
Vol 13 (16) ◽  
pp. 3278
Author(s):  
Arthur Moraux ◽  
Steven Dewitte ◽  
Bruno Cornelis ◽  
Adrian Munteanu
Keyword(s):  
Deep Learning ◽  
Satellite Imagery ◽  
Daily Rainfall ◽  
Radar Data ◽  
Rain Gauge ◽  
Precipitation Rate ◽  
Estimation Accuracy ◽  
Rate Estimation ◽  
The North ◽  
Precipitation Estimation

To improve precipitation estimation accuracy, new methods, which are able to merge different precipitation measurement modalities, are necessary. In this study, we propose a deep learning method to merge rain gauge measurements with a ground-based radar composite and thermal infrared satellite imagery. The proposed convolutional neural network, composed of an encoder–decoder architecture, performs a multiscale analysis of the three input modalities to estimate simultaneously the rainfall probability and the precipitation rate value with a spatial resolution of 2 km. The training of our model and its performance evaluation are carried out on a dataset spanning 5 years from 2015 to 2019 and covering Belgium, the Netherlands, Germany and the North Sea. Our results for instantaneous precipitation detection, instantaneous precipitation rate estimation, and for daily rainfall accumulation estimation show that the best accuracy is obtained for the model combining all three modalities. The ablation study, done to compare every possible combination of the three modalities, shows that the combination of rain gauges measurements with radar data allows for a considerable increase in the accuracy of the precipitation estimation, and the addition of satellite imagery provides precipitation estimates where rain gauge and radar coverage are lacking. We also show that our multi-modal model significantly improves performance compared to the European radar composite product provided by OPERA and the quasi gauge-adjusted radar product RADOLAN provided by the DWD for precipitation rate estimation.

Sign in / Sign up

Export Citation Format

Share Document