A neural network implementation for near real time tropospheric water vapour profiling over Nairobi using ground-based gps receiver
Abstract
A remote sensing tool employing an Artificial Neural Networks algorithm was proposed for
near real time determination of the relative humidity profile using Global Positioning System
(GPS) data recorded by a ground-based GPS receiver. The GPS data was processed to obtain
the Integrated Water Vapour. This Integrated Water Vapour in conjunction with ground level
information for temperature, pressure and relative humidity were fed as inputs to the
developed neural network which in turn generated the instantaneous relative humidity profile
as output.
GPS and radiosonde data for the years 2009 and 2010 were used to train the system while the
same data for 2011 were used to validate the system. The RH profile results for 2011
generated using GPS data and the neural network, upon comparison with recorded in situ
radiosonde relative humidity profile measurements for the same days and times in the year
2011, had Root Mean Square Error of less than 4%, which fell within the margin of error of
the Vaisala RS92 Radiosonde’s humidity measurement regime.
Publisher
University of Nairobi