We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Multi-step Lake Urmia water level forecasting using ensemble of bagging based tree models.
- Authors
Hajian, Rasul; Jalali, Mohammad Reza; Mastouri, Reza
- Abstract
Lakes play an important role in hydrologic cycle, and water level forecasting can provide vital information for future management of lakes and their ecosystem. In the present study, five standalone tree-based models including alternating model tree (AMT), M5 Prime (M5P), random forest (RF), random tree (RT) and reduced error pruning tree (REPT) with their ensemble with Bagging algorithm were applied to 1, 2, 3 and 6-months ahead water level forecasting at Lake Urmia, northwest of Iran. Water level time series data from 2001 to 2020 were obtained and divided into two sub-groups for model building (from 2001 to 2014) and validation (from 2015 to 2020). Different input scenarios were constructed to find the most effective input scenario. Finally developed models were evaluated through several visually-based and quantitative-based criteria. The obtained results indicated that for 1 and 2-months ahead, ensemble model of BA-AMT has a higher performance, while for 3 and 6-months ahead, BA-RF outperform other models. Overall, ensemble models enhanced forecasting power of standalone models. Also, developed models were able to forecast water level for up to 3-months ahead, and by increasing the scale for more months ahead, forecasting accuracy decreased and model were found to be of high uncertainty and low forecasting power.
- Subjects
IRAN; WATER levels; ECOSYSTEM management; BOOTSTRAP aggregation (Algorithms); LAKE management; HYDROLOGIC cycle; TREE pruning; FORECASTING
- Publication
Earth Science Informatics, 2022, Vol 15, Issue 4, p2515
- ISSN
1865-0473
- Publication type
Article
- DOI
10.1007/s12145-022-00857-w