
ABSTRACT Daily maximum temperature prediction plays an important role in climate‐risk management, agriculture, public health and energy planning and disaster preparedness in climate vulnerable country like Bangladesh. This study assesses the individual and ensemble predictive power of machine learning (ML), deep learning (DL) and hybrid ensemble‐based frameworks for daily maximum temperature prediction using 2013–2022 daily records from meteorological stations. Materials and methods Correlation‐based screening followed forward‐backward stepwise regression, standard scaling, a train‐test partitioning by chronology, and comparative validation with mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), Mean Absolute Percentage Error (MAPE) coefficient of determination () as performance measures. Four classical ML algorithms were evaluated such as Linear Regression, Random Forest, Support Vector Regression, and Extreme Gradient Boosting. Data driven recurrent DL models like Simple Recurrent Neural Network and Long Short Term Memory were trained with multivariate temporal windows. Three hybrid approaches were also tested: RNN‐LSTM sequential hybrid, stacking‐based ML ensemble, and combined DL‐ML super‐hybrid. The ML‐based models significantly outperformed standalone recurrent DL models in the given data configuration. XGBoost was also the best performing single ML model in terms of MAE = 0.76°C, MSE = 1.19°C 2 , RMSE = 1.09°C, MAPE = 3.21 and R 2 = 0.94. The ML stacking ensemble produced MAE = 0.76°C, MAPE = 3.18, MSE = 1.21°C 2 , RMSE = 1.05°C, and R 2 = 0.94. The best overall performance was obtained by using the proposed DL‐ML hybrid (MAE = 0.74°C, MAPE = 3.07, MSE = 1.08°C 2 , RMSE = 1.04°C, R 2 = 0.95). Our findings suggest that hybrid learning can enhance temperature prediction by integrating different types of meteorological relationships and representations in tabular and sequential forms. It also stresses, however, the requirement for independent multi‐station validation, leakage‐free stacking analysis, explain ability tests and extreme‐temperature conditions tests ahead of first deployment in the field.
ABSTRACT Since the beginning of the 21st century, global temperatures have exhibited a continuous upward trend, resulting in an increased frequency and intensity of extreme droughts and high‐temperature events worldwide. These climatic extremes have shown significant statistical associations with vegetation dynamics and terrestrial ecosystems. This study aimed to evaluate the response patterns of vegetation growth to extreme compound drought–heat events conditions on vegetation growth in the Poyang Lake Basin (PYLB). By integrating a conditional probability model with the Vine copula function, we quantitatively analysed the likelihood of vegetation growth inhibition under drought and heat stress during the growing seasons—from April to October—over the period 1982–2022. The main findings are as follows: (1) Vegetation in the central region of the basin showed the strongest response to drought, whereas vegetation in the northeastern and northwestern areas showed a more pronounced response to high‐temperature events. Notably, the central region exhibited the weakest response to heat stress. With increasing severity of drought and heat, the areas with stronger vegetation growth inhibition gradually shifted southward. (2) Vegetation in the Fuhe River Basin showed a distinct response pattern to drought conditions, whilst that in the Raohe River Basin was more responsive to elevated temperatures. (3) Cultivated vegetation showed the strongest growth inhibition response to both drought and compound drought–heat events, whereas shrublands exhibited a more sensitive response to high‐temperature stress.
ABSTRACT Ethiopia's economy, which depends heavily on agriculture, has been adversely affected by an increase in climate extreme events. Localized studies of climate extreme events of specific agroecological zones provides valuable insights to develop effective adaptation and mitigation strategies. The study focuses on analysis of rainfall and temperature extremes for the observed (1984–2014), midcentury (2030–2060) and end century (2070–2100) under SSP2‐4.5 & SSP5‐8.5 scenarios. Trend analysis was done using R statistical software Mann–Kendall package. Observed rainfall extremes, SDII, PRCPTOT, R10mm, R20mm and Rx1‐day rainfall showed significant increases at α < 0.05 to α < 0.01 in Wet Dega AEZs. PRCPTOT and R10mm increased significantly at α < 0.1 and α < 0.05, respectively, in Moist Kolla AEZs. R20mm and SDII showed significant trends at α < 0.05 and α < 0.01, respectively, in Moist Dega and Wet Weyna Dega AEZs. Rainfall extremes at the midcentury under SSP2‐4.5 and SSP5‐8.5 showed Rx10mm, Rx20mm, CWD and PRCPTOT) exhibited insignificant trends across AEZs. At the end century under SSP2‐4.5 and SSP5‐8.5 scenarios, PRCPTOT and RX10mm had significant increases ( α < 0.1) in some AEZs. The observed temperature extremes, TX10p and TN10p were significantly decreasing in Wet Dega AEZs at α < 0.1 and α < 0.001, respectively. TXx and TNn showed significant trends ( α < 0.05) in Moist Kolla AEZs. The projected temperature extremes at the midcentury under SSP2‐4.5 and SSP5‐8.5 showed that, TX10p, TX90p, TN10p, TN90p, SU, and TXx showed significant increases in most AEZs at α < 0.01 and α < 0.001, while, at the end century, TXn, TNx, TX90p, TN90p, SU and DTR exhibited insignificant increasing and significant increasing trends under SSP2‐4.5 and SSP5‐8.5 scenarios, respectively, across most AEZs. Thus, the findings highlight distinct vulnerability pathways among agroecological zones, offering critical insights for localized climate adaptation and mitigation strategies in Ethiopia.
Standard weather forecast evaluations focus on the forecaster's perspective and on a statistical assessment comparing forecasts and observations. In practice, however, forecasts are used to make decisions, so it seems natural to take the decision-maker's perspective and quantify the value of a forecast by its ability to improve decision-making. Decision calibration provides a novel framework for evaluating probabilistic forecast performance at the decision level rather than the forecast level. We evaluate decision calibration to compare a machine learning and a classical numerical weather prediction model on various weather-dependent decision tasks, though the framework is applicable to any set of forecast models. We find that model performance at the forecast level does not reliably translate to performance in downstream decision-making: some performance differences only become apparent at the decision level, and even among seemingly similar decision tasks, model rankings can change. Our results confirm that typical forecast evaluations are insufficient for selecting the optimal forecast model for a specific decision task.
ABSTRACT Sudden stratospheric warmings (SSWs) provide a source of surface predictability up to several weeks in advance. They generally precede cold weather across the northern midlatitudes, including extremes. When it is cold, demand for heating typically increases electricity demand. We use reanalysis data and a weather‐driven demand model for various European countries to investigate the relationship between SSWs and extreme electricity demand. We find that extreme demand of ~120% of climatology is 1.5–3 times more likely in periods following SSWs across northern and central Europe. We then use subseasonal forecast experiments from the Stratospheric Nudging and Predictable Surface Impacts project to explicitly attribute the impact of the February 2018 SSW, an event that preceded a cold weather outbreak across Europe. These experiments include nudging of the stratospheric state to either its observed or climatological evolution, allowing the causal impact of the SSW to be identified. We find a robust simulated increase in extreme demand risk attributed to the 2018 event. This indicates a window of opportunity for earlier prediction of extreme European electricity demand risks following SSWs.