One of the major drivers of ecosystem changes is fishing, with global wild fish harvests increasing four times from the 1950s to 1990s. This spike depleted fish biomass and altered ecosystems. To assess these economic and ecological impacts, projects such as the Fisheries and Marine Ecosystem Model Intercomparison Project (FishMIP) have been undertaken. By designing numerical experiments, FishMIP aims to clarify long-term climate impacts on fisheries and marine ecosystems, reducing uncertainties- in model projections. One of the models in FishMIP is the BiOeconomic mArine Trophic Size-spectrum (BOATS)- that integrates ecology and economics to simulate global fishing history. For the last round of FishMIP simulations focused on model evaluation (ISIMIP3a), I compared BOATS simulations forced by climate reanalysis and forced with historical fishing effort reconstructions to real-world observations. The goal of comparison is to help disentangle the complex interplay between fishing, climate change, and ecosystem health, to strengthen confidence in projections and ultimately aiding in sustainable fisheries management and marine conservation. For this comparison I have worked with two versions of BOATS, the former v1, and a new v2 that includes an improved representation of biodiversity. By analyzing these two versions, I address the following questions: how well the BOATS model performs compared to the observation? What are the effects of climate variability and the effect of human drivers? To investigate climate variability, I generated Empirical Orthogonal Function (EOFs) for the driving variables of BOATS, temperature, NPP, particle flux, and for the modeled consumer biomass. For human drivers, I developed and implemented diagnostics at Large Marine Ecosystems (LMEs) level, including correlation coefficient, root mean square, maximum catch and biomass variation that quantify the biomass decline in each LMEs. The analysis shows that the two model versions provide similar results, both globally and across LMEs. I highlighted coherent patterns of match/mismatch with observations that shed light on remaining model biases in particular at regional scales, and suggest future research directions.