This study analyzed time-series forecasting methods to predict maple sap harvesting for the production of maple syrup. These forecasting methods were evaluated with data from varied conditions collected by a sap harvesting remote monitoring device based on the Internet of Things (IoT) architecture. The primary problem this study investigated was the development of a predictive analytic capability to forecast the times when a sap collection tank level will be full and needs to be emptied. A full sap collection tank that is not emptied will result in a wasted overflow of sap, decreasing harvesting yield. An accurate forecast of tank level helps harvesters estimate when sap collection tanks will be full to better plan their workday. The forecasting methods explored included various baseline, statistical, and machine learning forecasting methods. A key requirement identified for accurate forecasts is the application of filtering and smoothing to remove outliers and reduce the variation of tank level data. This study found that a heuristic method based on simple drift forecasting provided good results for horizons out to 12 hours. This method is recommended for real-time forecasting of sap tank level as it provided high relative forecast accuracy and required significantly less computation time compared to more complex statistical and machine learning models.