Objective Photoacoustic tomography (PAT) holds significant potential for high-resolution deep-tissue imaging. In preclinical research, custom-designed concave arc-shaped ultrasound transducer arrays are often used to maximize the detection aperture. However, manufacturing limitations and assembly tolerances frequently cause the actual physical positions of array elements to deviate from their theoretical design. Additionally, concave arrays are typically covered with an acoustic lens, which introduces a mismatch in the speed of sound between the coupling medium and the lens material. The combination of these geometric and acoustic-phase errors leads to severe image artifacts, reduced contrast, and degraded resolution. This study proposes a systematic two-step calibration strategy to address these issues and substantially improve image quality.Methods First, a high-intensity isotropic photoacoustic point source was constructed using a multi-mode optical fiber coated with carbon nanotubes (CNTs) to acquire high signal-to-noise ratio calibration data. The Akaike information criterion (AIC) was employed to accurately determine the time of arrival (ToA) of photoacoustic signals. Subsequently, a geometric calibration algorithm based on nonlinear least-squares (NLS) estimation was developed. This algorithm iteratively solves for the true spatial coordinates of each array element by minimizing the residual between theoretical and measured acoustic path lengths. To further address sound-speed inhomogeneity caused by the acoustic lens, a phase compensation algorithm based on bilinear interpolation was proposed. This algorithm computes a pixel-specific phase delay map across the imaging region and performs point-by-point signal correction during delay-and-sum (DAS) reconstruction. The proposed methods were validated using a custom 96-channel concave arc-shaped array (center frequency: 12 MHz) through both phantom imaging and in vivo mouse tumor models.Results Phantom experiments showed that at an imaging depth of 14 mm, the reconstruction position deviation of the point source in the uncalibrated system reached up to 1 mm. After applying the combined calibration, the lateral resolution (full width at half maximum, FWHM) at the focal point of the arc array reached 95 μm—representing a 6.7-fold reduction compared to the uncalibrated state and a 4.8-fold reduction compared to geometric calibration alone without phase compensation. In vivo experiments demonstrated that the calibrated system clearly resolved the microvascular network of subcutaneous tumors in mice. Photoacoustic signals were strictly confined within tumor boundaries delineated by ultrasound imaging (USI), eliminating the vascular spillover artifacts commonly observed in uncalibrated images. Furthermore, after intravenous injection of indocyanine green (ICG), the system successfully detected weak photoacoustic signals at a depth of 5 mm, performing significantly better than the uncalibrated system.Conclusion The proposed calibration method, which integrates nonlinear least-squares estimation with phase compensation, significantly improves image fidelity and spatial resolution consistency across a wide field of view by correcting systemic geometric errors and acoustic phase aberrations. This approach demonstrates high robustness and provides a reliable technical foundation for the clinical translation of photoacoustic probes with non-standard geometries.