The Shelter Quality protocol (SQP) is a concise and easily implemented tool for assessing dog welfare and to identify critical aspects of the shelter environment. A first version of the protocol has been modified to improve its performance. The aim of this study was to evaluate the reliability of the measures included in the second version of the protocol (SQP2) by testing the interobserver agreement between two independent assessors. We evaluated the sensitivity of animal-based measures in detecting the shelter dogs' welfare outputs during two different seasons. Ten Italian shelters were assessed contemporaneously by two assessors to determine the reliability of SQP2 measures. Interobserver agreement was evaluated using the Cohen's kappa for qualitative variables and Pearson’ correlation for quantitative variables. The SQP2 was also applied twice (January and August) by the same observer in five Italian shelters to evaluate the sensitivity of the protocol to seasonal condition changes. The quantitative variables, “Number of animals shivering/huddling” and “Number of animals panting” were analyzed by Wilcoxon test. Credible intervals (95%) were calculated using a beta distribution for qualitative variables: “Body condition”, “Skin condition”, “Dog cleanliness”, “Signs of diarrhea”, “Coughing”, and “Lameness”. The level of agreement between the two observers on the qualitative variables such as body condition, lameness, skin condition, was quite high, ranging from substantial (0.61–0.80) to almost perfect (0.81–0.99). Interobserver agreement was also significant with Pearson correlation coefficients ranging from 0.51 to 0.92 (e.g., curious = 0.74; sociable = 0.83; barking level = 0.61). “Number of animals panting” and “Signs of diarrhea” showed a significant difference between the assessments (P<0.05). The observations of animals with lameness, coughing, and inadequate body condition increased in the winter season, whereas the observations of animals with skin lesions increased during the summer, but not significantly. The behaviors of shivering/huddling were observed too infrequently to be meaningfully analyzed. Consistent interobserver agreement exists in assessing dogs' welfare using the SQP2 confirming the reliability of the measures included in the protocol. The SQP2 shows potential in detecting changes in dogs' welfare outputs due to different climatic conditions. Further investigations are required to confirm the sensitivity of selected measures to different seasons.