Drought interval simulation using functional data analysis

Beyaztas U., Yaseen Z. M.

JOURNAL OF HYDROLOGY, cilt.579, 2019 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 579
  • Basım Tarihi: 2019
  • Doi Numarası: 10.1016/j.jhydrol.2019.124141
  • Dergi Adı: JOURNAL OF HYDROLOGY
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Anahtar Kelimeler: Drought, Functional data, Konya basin, PDSI, Time series, STANDARDIZED PRECIPITATION INDEX, NEURAL-NETWORK, SEVERITY INDEX, SPI, CLIMATE, MORTALITY, SELECTION, PACKAGE, MODELS, MARKOV
  • Marmara Üniversitesi Adresli: Hayır

Özet

Drought detection is an essential process for drought risk management and watershed sustainability. Describing a reliable predictive model for drought events has always been the motivation of the meteorology scientists. The current research proposes a functional time series analysis for the construction of a reliable predictive strategy of drought interval occurrences. The proposed model is validated with respect to its performance based on the Palmer Drought Severity Index (PDSI). The applied methodology is conducted to produce point and interval forecasts for the drought time series pattern. The modeling procedure is inspected on the Konya watershed, Central Anatolia, Turkey. The study is incorporated a second phase of modeling where entertained for heteroscedastic and seasonal hydro-climatic data (streamflow). Several statistical metrics and graphical presentation are computed for modeling evaluation. Two benchmark models are developed for modeling evaluation including linear exponential smoothing (LES) and autoregressive integrated moving average (ARIMA). In quantitative terms; for instance, the proposed functional principal component analysis (FPCA) model exhibited a remarkable drought interval prediction enhancement over the LES and ARIMA models by approximately 29.5% and 28.3% using the root mean square error (RMSE) metric at Aksaray station. In accordance with the attained results, the proposed methodology demonstrated a reliable predictive model for capturing future PDSI intervals and streamflow magnitudes.