Wireless Sensor Networks (WSN) are rapidly gaining importance for distributed sensing, control and automation tasks. Many of their applications require the individual nodes of the network to have their logical clocks running close to each other for reliable conveyance of information. Although a linear model of first order (i.e., involving the clock offset and skew) is enough in most cases, some applications such as efficient duty cycling and synchronized sampling require long term clock synchronization in the network which eventually increases its whole lifetime by spending lesser energy on communications for resynchronization. In this paper, we have proposed a quadratic model for better modeling the relationship between a node's clock with the reference and derived the Maximum Likelihood Estimator (MLE) for the corresponding clock drift, skew, offset and the fixed link delay assuming an exponential distribution for variable link delays. In addition, the algorithm to find such an estimator is also explained in detail.