Monthly precipitation assessments in association with atmospheric circulation indices by using tree-based models

Creative Commons License

Sattari M. T., Shaker Sureh F., Kahya E.

NATURAL HAZARDS, vol.102, pp.1077-1094, 2020 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 102
  • Publication Date: 2020
  • Doi Number: 10.1007/s11069-020-03946-5
  • Journal Name: NATURAL HAZARDS
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, IBZ Online, PASCAL, Aerospace Database, Agricultural & Environmental Science Database, Aquatic Science & Fisheries Abstracts (ASFA), CAB Abstracts, Communication Abstracts, Environment Index, Geobase, INSPEC, Metadex, PAIS International, Pollution Abstracts, Sociological abstracts, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Page Numbers: pp.1077-1094
  • Istanbul Technical University Affiliated: Yes


The Urmia Lake basin is one of the most important basins in Iran, facing many problems due to poor water management and rainfall reduction. Under current circumstances, it becomes critical to have an advanced understanding of rainfall patterns in the basin, setting the motivation of this study. In this research, the mean monthly meteorological data of six synoptic stations of Urmia Lake basin were used (including relative humidity, temperature, minimum-maximum temperature and pressure) and large-scale atmospheric circulation indices (Southern Oscillation Index, North Atlantic Oscillation, Western Mediterranean Oscillation, Mediterranean Oscillation-Gibraltar/Israel and Mediterranean Oscillation-Algiers/Cairo) and sea surface temperatures of the Mediterranean, Black, Caspian, Red seas and Persian Gulf in the period 1988-2016. Various combinations of these variables used as input to the M5 tree and random forest models were selected by Relief algorithm for each month in three scenarios including atmospheric circulation indices, meteorological variables and combination of both. After the implementation of two models with three different scenarios, the evaluation criteria including correlation coefficient (R), mean absolute error and root-mean-square error were calculated and the Taylor diagram for each model was plotted. Our results showed that the M5 tree model performed superior in January, February, March, April, June, September, November and December, while the random forest model did in the remaining months. In addition, the indications of this study showed that the combination of atmospheric circulation indices and meteorological variables used as input to the models mostly constituted improved results.