Public weather APIs provide wind forecasts accessible for pre-mission UAV trajectory planning. But systematic biases of 2−5m/s are commonly found in tropical coastal environments, directly impacting energy efficiency and safety. In this paper, a closed-loop system is proposed that incorporates IMU and airspeed sensor measurements with Bayesian forecast correction for real-time update of Conditional Value-at-Risk (CVaR) and trajectory planning. Using an Extended Kalman Filter (EKF), forecast bias and variance are addressed at each UAV waypoint. If the updated CVaR0.95 exceeds a threshold, warm start RRT* is utilized for trajectory planning within computational constraints. The proposed method is evaluated using a 1.94 km Sundarbans mission with synthetic wind data with known bias, achieving 72.0% RMSE reduction (2.04→0.57m/s), 12.6 % CVaR risk reduction, and 12.0 % energy savings versus the uncorrected baseline. EKF bias estimation achieves 79.6 % accuracy within five waypoints, ensuring reliable correction for the rest of the mission.