An improved representation of spatial precipitation over Türkiye using multiscale geographically weighted regression


DADAŞER ÇELİK F., KÖMÜŞCÜ A. Ü., ÇELİK M.

Stochastic Environmental Research and Risk Assessment, cilt.40, sa.11, 2026 (SCI-Expanded, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 40 Sayı: 11
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1007/s00477-026-03391-z
  • Dergi Adı: Stochastic Environmental Research and Risk Assessment
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, Environment Index, Geobase, Index Islamicus, zbMATH, Academic Search Ultimate (EBSCO), Natural Science Collection (ProQuest), Engineering Source (EBSCO), Materials Science & Engineering Collection (ProQuest), Pharma Collection (ProQuest), Technology Collection (ProQuest)
  • Anahtar Kelimeler: Interpolation, Multiscale geographically weighted regression, Precipitation
  • Erciyes Üniversitesi Adresli: Evet

Özet

Accurately estimating the spatial distribution of precipitation is essential for numerous applications that depend on distributed solutions, including hydrological modeling. This is typically achieved by applying interpolation methods, such as inverse distance weighted interpolation or kriging, to gauge-based precipitation data. However, these methods often struggle to capture spatial variability in regions with complex terrain fully. In this study, we propose using Multiscale Geographically Weighted Regression (MGWR) for spatial interpolation of precipitation. Although MGWR is an extension of Geographically Weighted Regression (GWR), its application to precipitation interpolation remains limited, and few studies have comprehensively compared its performance with conventional interpolation methods. Both GWR and MGWR consider spatially varying relationships between precipitation and explanatory variables, such as elevation and distance to the coast, to estimate values for unsampled locations. However, MGWR differs from GWR in that it incorporates multiple spatial scales, providing a better representation of spatial heterogeneity. We used precipitation data collected at 237 stations across Türkiye from 1980 to 2020, along with elevation and distance to coast, as explanatory variables. The performance of the MGWR was evaluated against several commonly used methods, including inverse distance weighted interpolation, local polynomial interpolation, Kriging, empirical Bayesian Kriging, co-kriging, and GWR. The results showed that the inclusion of explanatory variables considerably improved the results. MGWR and GWR outperformed the other methods in representing local variations in precipitation, with MGWR providing additional improvements by allowing a scale-dependent analysis. The MGWR also offered the lowest level of uncertainty. This study demonstrated that MGWR can potentially improve the spatial models of hydrological and climatic variables.