Abstract:
Radon-222 is a radioactive gas in the natural decay series of Uranium-238. It easily
emanates from the soil to constitute radiological hazard and is the leading cause of
lung cancer apart from smoking. High indoor radon buildup could occur in buildings
sited over high radon-bearing bedrocks. Radon hazard, expressed as Geogenic Radon
Potential (GRP), is due to a combination of soil-gas radon concentration ( ) and
soil-air permeability ( ), both of which depend on bedrocks. Data on these two
quantities over different bedrock formations and soil types in Southwest (SW) of
Nigeria are very scarce resulting in limited knowledge on radon hazard and lack of
requisite radon control guidelines. This study was designed to measure , determine
GRP and model the distribution of GRP over different bedrocks of SW Nigeria.
Measurements of were made using a calibrated real-time semiconductor radon
monitor at a depth of 0.80 – 1.00 m in 150 randomly selected locations across 20
bedrocks in the six states of SW Nigeria. Saturated hydraulic conductivities of
undisturbed soil samples taken from these locations were measured with a constanthead permeameter in order to determine . The GRP for each location was calculated
from and and categorised using Neznal classification for radon hazard ratings.
A Levenberg-Marquardt feed-forward-back-propagation artificial neural network was
employed to develop a predictive model for . Data was randomly split in 70:15:15
for training, testing and validation, respectively, for six different architectures and the
best was chosen following standard procedure. Goodness-of-Prediction (G), Average
Validation Error , Mean Bias Error and Root Mean Square Error
were used to determine performance and validation of the model. The and
GRP maps were generated on existing geological map for SW region.
The measured ranged . The ranged ,
while ranged . The GRP ranged .
Sedimentary formation had highest of , while granitic
bedrocks had highest and GRP of and ,
respectively. Radon hazard classification showed that , and of the sites
were of low, medium and high radon hazard rating, respectively. Out of the 13 sites
with high radon hazard rating, granitic and metamorphic bedrocks presented more sites
(84.6%). The best performing architecture was 2 x 8 x 1. Performance indices of the
model, yielded G of 73.5%, of 0.073, of 0.42 and of 4.62 kBqm-3.
Validation indices yielded G of 86 , of , of and of
1 , indicating good model performance. Values of measured and GRP
were used to generate maps which showed spatial distribution of low, medium and
high radon hazard ratings.
The values of measured soil-gas radon concentration and determined geogenic radon
potential were highest in granitic bedrocks. The performance indices of the developed
neural network model showed good reliability in predicting geogenic radon potential
for southwest Nigeria.
