All Time Past Year Past 30 Days
Abstract Views 1 0 0
Full Text Views 138 10 1
PDF Downloads 28 3 1

Estimating Tropical Pacific Rainfall Using Digital Satellite Data

Craig E. MotellDepartment of Land, Air and Water Resources, University of California, Davis, CA 95616

Search for other papers by Craig E. Motell in
Current site
Google Scholar
PubMed
Close
and
Bryan C. WeareDepartment of Land, Air and Water Resources, University of California, Davis, CA 95616

Search for other papers by Bryan C. Weare in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Statistical models that estimate tropical Pacific rainfall from the National Oceanic and Atmospheric Administration's global archive of polar-orbiter satellite data have been derived and tested. These rainfall models are based on the assumptions that rainfall is linearly related to bright visible and cold infrared radiation (IR) satellite. The models were derived by using measured monthly rainfall from small, flat, tropical islands with elevations less than 30 m together with digital IR and visible satellite data.

Three models were derived: one used visible and nighttime IR data (NIRVISQ); the second used only visible data (VISQ), and the third used an average of daytime and nighttime IR data (AVEIR). These models were found to predict between 62% and 67% of the variance of 1051 station-months of hindcast rainfall data measured from June 1974 through mid-March 1978 (J74M78). However, rainfall was found to be underpredicted on relatively high mean rainfall islands and vice versa. Similar prediction accuracies were found when the rainfall models were used to estimate rainfall on new low-latitude island stations during the J74M78 period. All three models showed a decrease in predictive skill during time periods after J74M78.

Tropical Pacific annual rainfall maps, estimated using the rainfall models and satellite data from June 1974 through May 1977, showed that NIRVISQ and VISQ may greatly overpredict rainfall in regions where stratus clouds are common such as in the eastern Pacific Ocean, but AVEIR appeared to predict reasonable rainfall amounts throughout the tropical Pacific. The AVEIR is thus the preferred model for predicting tropical oceanic rainfall.

Abstract

Statistical models that estimate tropical Pacific rainfall from the National Oceanic and Atmospheric Administration's global archive of polar-orbiter satellite data have been derived and tested. These rainfall models are based on the assumptions that rainfall is linearly related to bright visible and cold infrared radiation (IR) satellite. The models were derived by using measured monthly rainfall from small, flat, tropical islands with elevations less than 30 m together with digital IR and visible satellite data.

Three models were derived: one used visible and nighttime IR data (NIRVISQ); the second used only visible data (VISQ), and the third used an average of daytime and nighttime IR data (AVEIR). These models were found to predict between 62% and 67% of the variance of 1051 station-months of hindcast rainfall data measured from June 1974 through mid-March 1978 (J74M78). However, rainfall was found to be underpredicted on relatively high mean rainfall islands and vice versa. Similar prediction accuracies were found when the rainfall models were used to estimate rainfall on new low-latitude island stations during the J74M78 period. All three models showed a decrease in predictive skill during time periods after J74M78.

Tropical Pacific annual rainfall maps, estimated using the rainfall models and satellite data from June 1974 through May 1977, showed that NIRVISQ and VISQ may greatly overpredict rainfall in regions where stratus clouds are common such as in the eastern Pacific Ocean, but AVEIR appeared to predict reasonable rainfall amounts throughout the tropical Pacific. The AVEIR is thus the preferred model for predicting tropical oceanic rainfall.

Save