A cloud-based landslide and weather nowcasting system with collaborative machine learning for the Chitrakoot region in Mauritius

Landslides are a common geotechnical phenomenon which are influenced by weather conditions such as floods and particularly prevalent in tropical islands with volcanic domes such as Mauritius. The main purpose of this paper is to develop a cloud based real time landslide and weather forecasting system for the Chitrakoot region in Mauritius. The system consists of geotechnical sensors for land displacement and soil moisture measurements as well as weather sensors for rainfall, temperature, humidity, atmospheric pressure, wind direction and wind speed, deployed on two sites for collecting and transmitting data via a 4 G network to a cloud server. The cloud server monitors and displays the data in real-time via an interactive user interface allowing predictions to be performed for all the parameters being captured. The predictions are performed using the Multiple Linear Regression (MLR) and Multi-Layer-Perceptron algorithms with three different models that consider different correlations between the geotechnical and weather parameters to predict a parameter of interest. One of the models incorporates collaborative machine learning which combines weather and geotechnical parameters of both sites into the machine learning algorithms. The performance of the three models were analysed for both MLR and MLP with data collected from both sites for a period of one month. A Mean Absolute Percentage Error (MAPE) of as low as 0.02 % for land displacement predictions and 0.01 % for rainfall was achieved with the collaborative machine learning algorithm, under certain conditions. The main originality of this work is the development of a collaborative machine learning algorithm for a landslide and weather nowcasting system which integrates cloud computing and IoT.