Empirical Models For Estimating Precipitable Water Using Advanced Tiros Operational Vertical Sounder Satellite Data Over Peninsular Malaysia
Loading...
Date
2018-09
Authors
Makama Ezekiel Kaura
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Precipitable water (PW) is a highly variable, but important greenhouse gas that
regulates the radiation budget of the earth. Adequate knowledge on its distribution,
in space and time, is required for a better description and understanding of weather
and global climate. The few existing studies in Peninsular Malaysia utilized in situ
data to estimate TPW, with the consequences of oversimplifying the situation by
masking the effects of local circulation and topographical difference, which are
considered important in small scale weather studies. Also, the use of single
meteorological parameter to estimate TPW usually undermines other parameters that
may have combined effect on the column water vapour. New models based on
multiple linear regression (MLR) to estimate TPW have, therefore, been proposed
using homogenized climate data records derived from ATOVS onboard NOAA
along with surface observations for the period 2001 – 2011. ATOVS data agreed
well with radiosonde measurements, both spatially and seasonally, with correlation
coefficients (𝑟) ranging from 0.60 – 0.98. New relationship between lower layer
(WL) and other layers aloft, of the form 𝑊 = 𝛼(𝜑)[𝑊𝐿]𝛽(𝜑) have been proposed,
with W being either middle or higher layer PW, 𝛼 and 𝛽 are coefficients that are
functions of latitude (φ). The models gave 𝑟 and root mean square error (RMSE) of
respective values between 0.867 – 0.926 and 1.65 – 2.38 mm for both the WM and
WH predictions across the delineated zones. An overview of the spatial distribution of seasonal mean TPW showed general decreases from south to north over
Peninsular Malaysia, with intense horizontal gradients along the western boundary
during the NEM. Spatial and temporal variability of monthly mean of selected
meteorological parameters show that annual cycle and spatial pattern of relative
humidity are similar to those of TPW, with maximum values obtained in SWM and
least during the NEM period. Higher values are spatially found in the SZ with lower
values depicted in NZ. Temperature presents almost uniform structures across the
zones with maximum values in May/June and least in the month of January.
Generally, relative humidity had dominant spatial and temporal impact on TPW in
the entire study area. The proposed models gave very encouraging results in all the
zones. For the overall period, R2 was 0.967, 0.946, and 0.935 in NZ, CZ, and SZ
respectively, with MBE and RMSE being 0.09, 0.81, -0.97 mm and 0.93, 1.34, 1.68
mm in the same order. Artificial neural networks (ANN) models showed excellent
predictive power when contrasted with the MLR models. The performance of both
the MLR and ANN models vary in space and season, with best results in NZ and
least in SZ. They both showed better data fitting in most part of the NEM than the
SWM period. Generally, both the MLR and ANN exhibited great potentials in the
prediction of TPW in Peninsular Malaysia with the latter marginally outperforming
the former.
Description
Keywords
Precipitable water