Back to search

Aristotle University Of Thessaloniki (AUTH)

Συμβολή στην ανάπτυξη τεχνικής για την εκτίμηση της βροχόπτωσης από πολυφασματικά δορυφορικά δεδομένα

Abstract

dc:description

The present study aims at examining the potential of developing rainfall estimation schemes using the enhanced spectral resolution of the Meteosat Second Generation (MSG).Initially, the possibility of developing precipitating cloud detection schemes was investigated, using the enhanced thermal infrared spectral resolution of the Meteosat Second Generation (MSG). Two different classification methodologies were proposed that use the spectral parameters along with textural parameters derived from the thermal infrared MSG channels to delineate rain from no rain clouds. The first is an algorithm based on the probability of rain (Probability of Rainfall, PΟR) for each pixel of the thermal infrared MSG satellite data and the second is an Artificial Neural Network (Neural Network Μultilayer Perceptron, MLP) model. Both schemes were trained using as rain information spatially and temporally matched gauge data from 88 stations in Greece, for 30 rainy cases covering the period from March 2008 to February 2009. Both schemes were evaluated against an independent sample of rain gauge data for ten rainy days. During the training phase, POR1 model based on spectral parameters showed the best performance among all the rain area delineation models while MLP2 model exhibited the lowest performance. When evaluating against the independent dataset, the MLP1 model provides the best results among all the rain area discrimination techniques and POR2 algorithm produces the worst results. From the validation results, it was found that the introduction of textural parameters does not improve the discrimination between rain and no rain clouds. All algorithms overestimate the rain occurrences detected by the rain stations network.The next step of the rainfall estimation methodology was the development of two convective and stratiform rain delineation schemes based on the high spectral resolution of the MSG. Two different classification methods were proposed that use spectral cloud parameters along with textural cloud parameters. The first model is an empirical algorithm based on the estimation of the probability of convective rainfall on a pixel basis (Probability of Convective Rainfall, PCR) for the satellite infrared dataset and the second is an Artificial Neural Network model (MLP) for convective/stratiform rain classification that is based on the correlation between spectral and textural parameters and convective and stratiform classes of surface rainfall. The rain delineation algorithms were trained using different rainy cases with high convective activity covering the same period from March 2008 to February 2009. The PCR2 algorithm that is based on both spectral and textural measures, during the training phase provided the best results among all the rain classification models and MLP1 algorithm exhibited the worst scores. When evaluating against the independent dataset, the MLP2 model based on both spectral and textural parameters showed the best performance among all the convective/stratiform rain discrimination schemes and algorithm PCR1 provided the lowest performance. The verification results showed that the incorporation of textural parameters as additional information has improved the convective/stratiform rain discrimination compared to the use of spectral information alone. All algorithms overestimate the convective rain occurrences observed by the rain gauge network.The main objective of the present study was the development of two multispectral rainfall estimation techniques based initially on the developed rain and non rain discrimination models and then on the convective/stratiform rain classification models that incorporate the enhanced spectral resolution of the Meteosat Second Generation. Two different rainfall estimation methods were introduced that use spectral cloud parameters along with textural cloud parameters. The first is an algorithm based on rainfall rate assignment (Rainfall Rate assignment, RR) for each pixel of the thermal infrared MSG satellite images and the second is an Artificial Neural Network MLP algorithm. The performance of the rainfall estimation schemes was evaluated at different temporal averaging intervals (15min, 1h, 3h, 6h).The intercomparison between the two rainfall estimation techniques showed that rainfall retrieval algorithms based on the convective and stratiform rain delineation model improve rainfall estimation technique compared to algorithms that rely on the rain cloud delineation model. The rainfall estimation algorithm MLP2, which was developed according to the convective/stratiform classification model, provided the best results among all the rainfall retrieval algorithms while RR1 algorithm produced the worst results for each time interval. Furthermore between the two rainfall rate assignment algorithms, the RR2 algorithm displayed the best scores compared with the RR1 algorithm. The evaluation of rainfall estimation algorithms that were developed based on the rain and non rain delineation models exhibited the lowest performance at the first time interval of 15min and then was significantly improved for the next temporal averaging intervals (1h, 3h, 6h).The validation of rainfall retrieval algorithms at independent rainy days displayed that rainfall estimation algorithms based on the convective and stratiform rain classification model provide better results compared to algorithms that rely on the rain cloud detection model.As a final output, a multispectral rainfall estimation algorithm was developed based on the MSG satellite data and gauge rainfall measurements that generate precipitation estimates at various temporal scales.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2013

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Giannakos, Apostolos
  • Γιαννακός, Απόστολος

Subjects

dc:subject × 21

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/35628

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Giannakos, Apostolos; Γιαννακός, Απόστολος. Συμβολή στην ανάπτυξη τεχνικής για την εκτίμηση της βροχόπτωσης από πολυφασματικά δορυφορικά δεδομένα. Aristotle University Of Thessaloniki (AUTH), 2013. http://hdl.handle.net/10442/hedi/35628