Recent times, the Wireless Sensor Networks (WSN) has played an important role in smart farming systems. However, WSN-enabled smart farming (SF) systems need reliable communication to minimize overhead, end-to-end delay, latency etc., Hence, this work introduces a 3-tiered framework based on the integration of WSN with the edge and cloud computing platforms to acquire, process and store useful soil data from agricultural lands. Initially, the sensors are deployed randomly throughout the network region to collect information regarding different types of soil components. The sensors are clustered based on distance using the Levy flight based K-means clustering algorithm to promote efficient communication. The Tasmanian devil optimization (TDO) algorithm is used to choose the cluster heads (CHs) based on the distance among the node and edge server, residual energy, and the number of neighbors. Then, the optimal paths to transmit the data are identified using the all members group search optimization (AMGSO) algorithm based on different parameters. Each edge server assesses the quality of the data (QoD) with respect to some data quality criteria after receiving the data from the edge server. Also, the load across the servers are balanced in order to overcome the overloading and under loading issues. The legitimate data that received higher scores in the QoD evaluation alone is sent to the cloud servers for archival. Using the ICRISAT dataset, the efficiency of the proposed work is evaluated using a number of indicators. The average improvement rate attained by the proposed model in terms of energy consumption is 40%, in terms of packet delivery ratio is 7%, in terms of network lifetime is 38%, and in terms of latency is 24% for a total of 250 nodes.