Scholarlink

<p>Projecting precipitation changes through a multi-model ensemble (MME) of global climate models (GCMs) enhances predictive accuracy by reducing model uncertainty, thereby offering more reliable insights for adaptive water resource management under future climate scenarios. Initially, grid-wise ranking of 27 NASA NEX-GDDP-CMIP6 GCMs is performed by using four performance indicators for which weights are assigned by using the entropy method. These weighted performance indicators are then given as input to the technique for order preference by similarity to ideal solution method to derive GCM rankings at each grid point over the Ujjani Dam. The group decision-making approach (GDMA) is utilized to aggregate the rankings of 57 grid points over the Ujjani Dam. MMEs are created at each grid point by using top 50% GCMs obtained from the GDMA by employing extreme gradient boosting (XGBOOST), support vector machine, reliability ensemble averaging, adaptive boosting (AdaBoost), random forest (RF), arithmetic mean, and multiple linear regression (MLR). The findings of the current study revealed that (1) the top 5 ranked GCMs determined by GDMA are MIROC-ES2L, GFDL-CM4-gr2, MPI-ESM1-2-LR, KIOST-ESM, and BCC-CSM2-MR. (2) All developed MMEs are found to perform better than the top-ranked individual GCMs. (3) MMEs of RF, XGBOOST, AdaBoost, and MLR have yielded better results and shown good agreement with observed data. (4) Projected precipitation changes for future periods are as follows: near future under SSP245 from&#xa0;− 46.9% to 41%&#xa0;and under SSP585&#xa0;from − 48.7% to 41.5%; mid-future under SSP245&#xa0;from − 41.1% to 57.9%&#xa0;and under SSP585&#xa0;from − 37.8% to 88.5%; far future under SSP245&#xa0;from − 37.2% to 80.92%&#xa0;and under SSP585&#xa0;from − 31.7% to 179.5%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ranking of CMIP6 GCMs and formulation of multi-model ensembles for precipitation projection under SSP245 and SSP585 scenarios over the Ujjani Dam catchment in India by using machine learning and conventional methods

  • 期刊论文

  • 发表日期：

    2025年7月1日

查看全文
  • Jarpala Venkatesh,
  • Ganesh D. Kale

摘要

Projecting precipitation changes through a multi-model ensemble (MME) of global climate models (GCMs) enhances predictive accuracy by reducing model uncertainty, thereby offering more reliable insights for adaptive water resource management under future climate scenarios. Initially, grid-wise ranking of 27 NASA NEX-GDDP-CMIP6 GCMs is performed by using four performance indicators for which weights are assigned by using the entropy method. These weighted performance indicators are then given as input to the technique for order preference by similarity to ideal solution method to derive GCM rankings at each grid point over the Ujjani Dam. The group decision-making approach (GDMA) is utilized to aggregate the rankings of 57 grid points over the Ujjani Dam. MMEs are created at each grid point by using top 50% GCMs obtained from the GDMA by employing extreme gradient boosting (XGBOOST), support vector machine, reliability ensemble averaging, adaptive boosting (AdaBoost), random forest (RF), arithmetic mean, and multiple linear regression (MLR). The findings of the current study revealed that (1) the top 5 ranked GCMs determined by GDMA are MIROC-ES2L, GFDL-CM4-gr2, MPI-ESM1-2-LR, KIOST-ESM, and BCC-CSM2-MR. (2) All developed MMEs are found to perform better than the top-ranked individual GCMs. (3) MMEs of RF, XGBOOST, AdaBoost, and MLR have yielded better results and shown good agreement with observed data. (4) Projected precipitation changes for future periods are as follows: near future under SSP245 from − 46.9% to 41% and under SSP585 from − 48.7% to 41.5%; mid-future under SSP245 from − 41.1% to 57.9% and under SSP585 from − 37.8% to 88.5%; far future under SSP245 from − 37.2% to 80.92% and under SSP585 from − 31.7% to 179.5%.