Offshore wind has an important role in achieving net zero targets and is a vital contributor to addressing the energy crisis. Continued expansion in offshore wind is driving industry and governments across the world to explore deeper and more exposed sites. Standard wind measurements using meteorological masts are more challenging, or even impractical at such sites. As a result, floating LIDAR and physics-based mesoscale models are increasingly being deployed as cost-effective solutions for resource assessment. This study investigates the performance of a meso-microscale coupled Weather Research and Forecasting model by comparing against using 15-months of in-situ wind data collected from two floating LiDAR systems deployed in the Celtic Sea, Southwest UK. Wind speed measurements from model and measurement were compared directly. Vertical wind speed gradient estimates, chosen to quantify how the two data sets predicted wind shear, were also compared, along with estimates of energy generation, as predicted by the two data sets. At approximate turbine hub height (150 m), the model predicted mean wind speeds are 2.38% and 2.81% lower than the measurements and 4.96% and 4.51% lower at the P95 wind speed, for the two measurement locations. However, the differences are greater at wind speeds faster than the turbine cut-out speed. The inverse is true at for windspeeds, below the turbine cut-in speed, where model predicted wind speeds are greater than the measurements. As these high and low wind speeds do not contribute to energy generation, the differences between model-predicted and measurement-predicted energy production are smaller, with model estimates lower by 0.56% and 1.32% for the two measurement sites. Model calculated wind speed gradients were notably different from the measurement-calculated wind speed gradients. In particular, the 90 ^th percentile value is 35% lower than that predicted by measurements. Attempts to account for this when calculating energy generation suggested that the model estimates were lower than measurements by a much lower margin of 0.6% and 1.37%for the two floating LiDAR locations. This work identifies differences between numerical model predictions and floating Lidar measurements for a proposed offshore wind site. Of the parameters considered, the energy production estimates show the least difference. This helps inform studies looking to accurately evaluate wind resource at more exposed sites. However, usage of wind data that specifically relies on accuracy for large or small wind speeds, such as engineering design, or weather window calculations will be required to consider the impact of the observed differences on their results.