Lightning disasters are now recognised as one of the top ten most severe natural calamities, being particularly frequent in mountainous areas. The continuous growth of tourism has led to significant impacts on tourists and cable cars, especially in mountainous scenic areas, where equipment like cable cars are highly sensitive to lightning. To investigate the key factors influencing lightning development in these regions and to promptly understand the trends in lightning activity in and around the Huangshan Scenic Area, we are leveraging multiple monitoring data, such as Doppler weather radar, meteorological soundings, and lightning detection. In this study, we are building and evaluating multiple lightning nowcasting models using different machine learning algorithms. The models are based on the non-inductive charging mechanism in the thunderstorm and the characteristics of Doppler weather radar echo. We are extracting echo characteristics of the Doppler weather radar as key forecasting factors, focusing on the intensity, vigour, and movement trends of the thunderstorm system. By comparing false alarm rates, missed alarm rates, and TS scores of various machine learning algorithms, we are selecting the most suitable forecast method for mountainous scenic areas. Our evaluation results reveal that the Random Forest (RF), Logistic Regression (LR), K-Nearest Neighbour (KNN), Gaussian Naive Bayes (GNB), and Support Vector Machine (SVM) algorithms all have certain nowcasting capabilities for lightning. The RF algorithm scores highest in TS scoring, the SVM has the lowest missed alarm rate, and LR has the lowest false alarm rate. Among these, the intensity and vigorous development of the thunderstorm system play a pivotal role in the RF algorithm, with the radar base reflectivity at the -20 ℃ layer height in the thunderstorm system intensity having the most influence, followed by the radar echo thickness above the 0 ℃ layer. Taking an example on 29 August 2021, when a large-scale intense thunderstorm occurred in and around the Huangshan Scenic Area in the afternoon. Employing the RF method resulted in a false alarm rate of 0.425, a missed alarm rate of 0.378, and a TS score of 0.426, indicating good forecasting performance in the area.