Biodiversity plays an important role in maintaining a healthy ecosystem. However, monitoring biodiversity in a noninvasive manner can be challenging. A potential solution for this, at least in the case of avian species, is monitoring through audio analysis. By utilizing audio analysis on field recordings, bird species could be identified by their call in a minimally invasive way. To support this idea, Mel frequency cepstral coefficients (MFCCs) are explored for audio encoding in combination with convolutional neural networks (CNNs) to efficiently process audio and identify specific spaces despite the background and ambient noise. To ensure favorable performance, a modified metaheuristic optimizer, a variation of the baseline reptile search algorithm (RSA), is introduced to handle hyperparameter tuning. To validate this approach, a publicly available dataset comprised of real-world bird field audio recordings is encoded using MFCC, and classification is handled by optimized CNN. Optimization is conducted as well as a comparative analysis between several state-of-the-art optimization algorithms and the proposed modified metaheuristic. The introduced algorithm attained the best performance with an accuracy of 0.838710.