This paper proposes a novel approach called the Mel_Graph-GCN framework, which utilizes graph convolutional neural networks to identify multiple bird species from field recordings. The process involves creating a graph from the Mel-spectrogram of the audio file using a trained deep convolutional neural network (deep CNN), and employing SpecAugment to generate additional Mel-spectrograms for enhanced training of the deep CNN. Subsequently, the graph is fed to a GCN for classification. The algorithm's performance is evaluated using the Xeno-canto bird sound database and compared with state-of-the-art models, demonstrating superior performance with a macro F1 score of 0.85.