A paradigm of novel-networking is presented by the Underwater Wireless Sensor Networks (UWSNs) when compared to Terrestrial Wireless Sensor Networks. Its not straightforward, instead, basic challenges need to be addressed for the deployment of UWSNs due to the environment type found underwater. UWSNs have to depend on other physical means such as acoustic signals for the transmission as the electromagnetic waves cannot be transmitted over a long distance in underwater environment. Large latency and low bandwidth are the key features of underwater wireless link as compared to the wireless link among ground-based sensors. The nature of transmission medium and physical properties of the environment of underwater acoustic channels are temporally and spatially variable. High variations occurring in underwater acoustic channels result in high uncertainties to precisely model the signal attenuation which is dependent on transmission link length and frequency. This paper has been intended to address such type of uncertainties and closely examine even minor variations occurring in signal attenuation in cases of spherical and cylindrical spreading. These variations have been addressed by using a mathematical modeling technique as `Gradient Estimation Vector'. It is the technique for systematically changing parameters in a model to determine the effects of such changes. Gradient Estimation Vectors actually characterize the signal attenuation more precisely along with the variations and uncertainties involved.