METEOSAT-8 provides new means to address the challenge of using geostationary satellite data to estimate precipitation: the SEVIRI sensor has higher spatial and temporal resolution and a higher number of spectral channels than the old METEOSAT sensor. This work aims to assess quantitatively the potential improvement due to the new channels. Appropriate visible (VIS) and infrared (IR) channels on the MODIS instrument have been used to simulate SEVIRI channels. Artificial Neural Network (ANN) techniques have been adopted to establish an indirect procedure to estimate Probability of Precipitation (PoP). The PoP ANN estimator is used to classify cloudy pixels into rain and no rain classes (with a threshold value of 1/32 mm h-1) and its performance has been assessed in terms of the Equitable Threat Score (ETS). The analysis involves winter season and morning/early night hours. The results obtained were compared with the most nearly corresponding METEOSAT-based scheme available in the Nimrod nowcasting system of the Met Office (UK). The evident impact of new channels on the performance shows that a new SEVIRIbased scheme could provide better estimation of precipitation. An optimum set of channels and features (Local Variability and Textural) useful for an operational operative PoP estimator has been selected and proposed in both day and night-time cases. In daytime, the 1.6μm channel combined with a visible channel showed remarkable skill at distinguishing raining from non-raining pixels.
Probability of precipitation using SEVIRI-like data and artificial neural networks.
PORCU', Federico;
2004
Abstract
METEOSAT-8 provides new means to address the challenge of using geostationary satellite data to estimate precipitation: the SEVIRI sensor has higher spatial and temporal resolution and a higher number of spectral channels than the old METEOSAT sensor. This work aims to assess quantitatively the potential improvement due to the new channels. Appropriate visible (VIS) and infrared (IR) channels on the MODIS instrument have been used to simulate SEVIRI channels. Artificial Neural Network (ANN) techniques have been adopted to establish an indirect procedure to estimate Probability of Precipitation (PoP). The PoP ANN estimator is used to classify cloudy pixels into rain and no rain classes (with a threshold value of 1/32 mm h-1) and its performance has been assessed in terms of the Equitable Threat Score (ETS). The analysis involves winter season and morning/early night hours. The results obtained were compared with the most nearly corresponding METEOSAT-based scheme available in the Nimrod nowcasting system of the Met Office (UK). The evident impact of new channels on the performance shows that a new SEVIRIbased scheme could provide better estimation of precipitation. An optimum set of channels and features (Local Variability and Textural) useful for an operational operative PoP estimator has been selected and proposed in both day and night-time cases. In daytime, the 1.6μm channel combined with a visible channel showed remarkable skill at distinguishing raining from non-raining pixels.I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
