ASAddisu Gezahegn Semie

Addisu Gezahegn Semie

*This is an algorithmically generated author record based on publicly available information.

Addisu Gezahegn Semie has a total of 23 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Inter-Comparison of Deep Learning Models for Flood Forecasting in Ethiopia’s Upper Awash Basin

    by Girma Moges Mengistu, Addisu G. Semie, Gulilat T. Diro, Natei Ermias Benti, Emiola O. Gbobaniyi, Yonas Mersha

    Flood events driven by climate variability and change pose significant risks for socio-economic activities in the Awash Basin, necessitating advanced forecasting tools. This study benchmarks five deep learning (DL) architectures, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Bidirectional LSTM (BiLSTM), and a Hybrid CNN–LSTM, for daily discharge forecasting for the Hombole catchment in the Upper Awash Basin (UAB) using 40 years of hydrometeorological observations (1981–2020). Rainfall, lagged discharge, and seasonal indicators were used as predictors. Model performance was evaluated against two baseline approaches, a conceptual HBV rainfall–runoff model as well as a climatology, using standard and hydrological metrics. Of the two baselines (climatology and HBV), the climatology showed limited skill with large bias and negative NSE, whereas the HBV model achieved moderate skill (NSE = 0.64 and KGE = 0.82). In contrast, all DL models substantially improved predictive performance, achieving test NSE values above 0.83 and low overall bias. Among them, the Hybrid CNN–LSTM provided the most balanced performance, combining local temporal feature extraction with long-term memory and yielding stable efficiency (NSE ≈ 0.84, KGE ≈ 0.90, and PBIAS ≈ −2%) across flow regimes. The LSTM and GRU models performed comparably, offering strong temporal learning and robust daily predictions, while BiLSTM improved flood timing through bidirectional sequence modeling. The CNN captured short-term variability effectively but showed weaker representation of extreme peaks. Analysis of peak-flow metrics revealed systematic underestimation of extreme discharge magnitudes across all models. However, a post-processing flow-regime classification based on discharge quantiles demonstrated high extreme-event detection skill, with deep learning models exceeding 89% accuracy in identifying extreme-flow occurrences on the test set. These findings indicate that, while magnitude errors remain for rare floods, DL models reliably discriminate flood regimes relevant for early warning. Overall, the results show that deep learning models provide clear improvements over climatology and conceptual baselines for daily streamflow forecasting in the UAB, while highlighting remaining challenges in peak-flow magnitude prediction. The study indicates promising results for the integration of deep learning methods into flood early-warning workflows; however, these results could be further improved by adopting a probabilistic forecasting framework that accounts for model uncertainty.


  • Open Access

    Towards Improved Flash Flood Forecasting over Dire Dawa, Ethiopia Using WRF-Hydro

    by Addisu G. Semie, Gulilat T. Diro, Teferi Demissie, Yonas M. Yigezu, Binyam Hailu

    Flash floods are increasingly frequent worldwide. Recent flooding in eastern Ethiopia resulted in casualties, destruction of property and interruptions of service. National flash flood forecasts made today primarily consider precipitation, putting less emphasis on surface processes. Enhancing accurate flash flood forecasts by accounting for surface processes and hydrological models together with a deeper understanding of heavy precipitation mechanisms is of paramount importance. To this end, an uncoupled WRF-Hydro model was calibrated for eastern Ethiopia to simulate extreme floods. Sensitivity analysis for August 2006 showed that infiltration runoff, hydraulic soil conductivity and saturated volumetric soil moisture with parameter values of 0.1, 1.5 and 1.0 produced realistic streamflow distribution. Extreme floods in March 2005 and April 2007 were further studied. The results showed that WRF-Hydro replicates temporal and spatial patterns well. Analysis using observational/reanalysis data revealed associated physical processes. Precipitation during these events exceeded long-term climatology and spanned wider areas in eastern Ethiopia. These heavy precipitation events are associated with strong upper-level westerly jet streams and rainfall-conducive circulation anomalies at lower levels. Positive outcomes from WRF-Hydro suggest operational implementation for flood monitoring and early warning systems in forecasting centers.


  • Open Access

    The Influence of Convective Aggregation on the Stable Isotopic Composition of Water Vapor

    by Joseph Galewsky, Matthias Schneider, Christopher Diekmann, Addisu Semie, Sandrine Bony, Camille Risi, Kerry Emanuel, Helene Brogniez

    Remote sensing datasets of water vapor isotopic composition are used along with objective measures of convective aggregation to better understand the impact of convective aggregation on the atmospheric hydrologic cycle in the global tropics (30°N to 30°S) for the period 2015–2020. When convection is unaggregated, vertical velocity profiles are top‐heavy, mixing ratios increase and water vapor δD decreases as the mean precipitation rate increases, consistent with partial hydrometeor evaporation below anvils into a relatively humid atmospheric column. Aggregated convection is associated with bottom‐heavy vertical velocity profiles and a positive correlation between mixing ratio and δD, a result that is consistent with isotopic enrichment from detrainment of shallow convection near the observation level. Intermediate degrees of aggregation do not display significant variation in δD with mixing ratio or precipitation rate. Convective aggregation provides a useful paradigm for understanding the relationships between mixing ratio and isotopic composition across a range of convective settings. The results presented here may have utility for a variety of applications including the interpretation of paleoclimate archives and the evaluation of numerical simulations of convection.


  • Open Access

    Estimating solar radiation using artificial neural networks: a case study of Fiche, Oromia, Ethiopia

    by Tegenu Argaw Woldegiyorgis, Natei Ermias Benti, Mesfin Diro Chaka, Addisu Gezahegn Semie, Ashenafi Admasu Jemberie


  • Open Access

    Towards Improved Flash Flood Forecasting Using WRF-Hydro in the Horn of Africa: Case of Dire Dawa, Ethiopia

    by Addisu Gezahegn Semie, Gulilat Tefera Diro, Teferi D Demissie, Yonas Mersha Yigezu, Binyam Tesfaw Hailu

    A reliable flood early warning system must take into account the mechanisms that cause heavy precipitation events and accurate surface hydrology modeling. In this project, analysis of atmospheric processes and hydrological modelling of selected flood events over Dire Dawa is conducted using various observational/reanalysis data and uncoupled WRF-Hydro model simulations. To comprehend the processes causing such severe precipitation occurrences, large scale atmospheric fields linked to selected extreme precipitation events are examined using ERA5 reanalysis. The land surface was configured at 1 km resolution while 250 m sub-grid resolution was set to perform the routing process. Model forcing for the uncoupled WRF-Hydo model is obtained from ERA5 reanalysis data. Sensitivity of stream-flow simulation to various parameter values such as hydrolic conductivity and surface infiltration coefficient was carried out for August 2006. The result of the sensitivity experiment reveals that infiltration-runoff, hydrolic soil conductivity and saturated volumetric soil moisture with the parameter value of 0.1, 1.5 and 1.0, respectively are found to produce realistic spatial and temporal distribution of stream-flow. The extreme flood events of March 2005 and April 2007 were studied further to assess the performance of WRF-Hydro model and to understand the underlying atmospheric mechanisms causing these heavy precipitation events. The result of hydrological simulation demonstrated that uncoupled WRF-Hydro simulation reproduced both the temporal evolution and the spatial pattern reasonably well. Our analysis indicated that the amount of precipitation during these two events exceeded the long-term average by several factors, furthermore, the anomalies cover larger areas of eastern Ethiopia. Associated to these extreme events, upper level subtropical westerly jet-streams were anomalously stronger and also extended further southward favouring upper level divergence over the region. At lower level, the notable circulation anomalies include anomalous positive pressure anomaly over Sudan/Egypt leading to northerly flow anomaly over Red Sea, strengthening of southerly influx from southern Indian ocean due to stronger Mascarene High. The encouraging results from WRF-Hydro simulation suggest that this modelling framework can be implemented in operational context within national and regional forecasting centers as a key component to establish a flood monitoring and early warning system.

  • Top Cited Papers
    Open Access

    Forecasting Renewable Energy Generation with Machine Learning and Deep Learning: Current Advances and Future Prospects

    by Natei Ermias Benti, Mesfin Diro Chaka, Addisu Gezahegn Semie

    This article presents a review of current advances and prospects in the field of forecasting renewable energy generation using machine learning (ML) and deep learning (DL) techniques. With the increasing penetration of renewable energy sources (RES) into the electricity grid, accurate forecasting of their generation becomes crucial for efficient grid operation and energy management. Traditional forecasting methods have limitations, and thus ML and DL algorithms have gained popularity due to their ability to learn complex relationships from data and provide accurate predictions. This paper reviews the different approaches and models that have been used for renewable energy forecasting and discusses their strengths and limitations. It also highlights the challenges and future research directions in the field, such as dealing with uncertainty and variability in renewable energy generation, data availability, and model interpretability. Finally, this paper emphasizes the importance of developing robust and accurate renewable energy forecasting models to enable the integration of RES into the electricity grid and facilitate the transition towards a sustainable energy future.


  • Open Access

    Forecasting Renewable Energy Generation with Machine learning and Deep Learning: Current Advances and Future Prospects

    by Natei Ermias Benti, Mesfin Diro Chaka, Addisu Gezahegn Semie

    This article presents a review of current advances and future prospects in the field of forecasting renewable energy generation using machine learning (ML) and deep learning (DL) techniques. With the increasing penetration of renewable energy sources (RES) into the electricity grid, accurate forecasting of their generation becomes crucial for efficient grid operation and energy management. Traditional forecasting methods have limitations, and thus ML and DL algorithms have gained popularity due to their ability to learn complex relationships from data and provide accurate predictions. This paper reviews the different approaches and models that have been used for renewable energy forecasting and discusses their strengths and limitations. It also highlights the challenges and future research directions in the field, such as dealing with uncertainty and variability in renewable energy generation, data availability, and model interpretability. Finally, this paper emphasizes the importance of developing robust and accurate renewable energy forecasting models to enable the integration of RES into the electricity grid and facilitate the transition towards a sustainable energy future.


  • Open Access

    The influence of convective aggregation on the stable isotopic composition of water vapor: Implications for the humidity of the troposphere

    by Joseph Galewsky, Matthias Schneider, Christopher Johannes Diekmann, Addisu Gezahegn Semie, Sandrine Bony, Camille Risi, Kerry A. Emanuel, Hélène Brogniez

    Remote sensing datasets of water vapor isotopic composition are used along with objective measures of convective aggregation to better understand the impact of convective aggregation on the atmospheric hydrologic cycle in the global tropics ($30^{\circ}$N to $30^{\circ}$S) for the period 2015-2020. When convection is unaggregated, vertical velocity profiles are top-heavy, mixing ratios increase and water vapor $\delta D$ decreases as the mean precipitation rate increases, consistent with partial hydrometeor evaporation below anvils into a relatively humid atmospheric column. Aggregated convection is associated with bottom-heavy vertical velocity profiles and a positive correlation between mixing ratio and $\delta D$, a result that is consistent with isotopic enrichment from detrainment of shallow convection near the observation level. Intermediate degrees of aggregation do not display significant variation in $\delta D$ with mixing ratio or precipitation rate. Convective aggregation provides a useful paradigm for understanding the relationships between mixing ratio and isotopic composition across a range of convective settings. The results presented here may have utility for a variety of applications including the interpretation of paleoclimate archives and the evaluation of numerical simulations of convection.

  • Open Access

    Impact of a Mixed Ocean Layer and the Diurnal Cycle on Convective Aggregation

    by Adrian M. Tompkins, Addisu G. Semie

    We investigate how ocean feedbacks and the diurnal cycle impact convective aggregation using a slab ocean coupled to a cloud resolving model. With a 20 m mixed layer ocean, aggregation occurs after 25 days. Thinner ocean layers slow the onset of clustering, with a 1 m ocean layer needing around 43 days. The delay is due to anomalous solar radiation in clear sky regions, causing a relative surface warming balanced by a matching cooling in cloudy areas. The resulting gradient in surface heat fluxes opposes low level convergence into convecting regions. On aggregation, the convective regions are surrounded by moist, clear sky regions with the hottest SSTs, toward which convection migrates, while a cold SST patch forms under dry suppressed regions due to enhanced latent heat fluxes and longwave radiation. The next experiment allows a diurnal cycle of 2.5°C in the domain mean SST. This causes convective rainfall to shift from a weak morning maximum to a sharper evening peak, reminiscent of undisturbed tropical observations. However, convection reverts to a weak early morning maximum once aggregation starts due to spatially heterogeneous radiative forcing. This implies thin mixed ocean layers are a necessary but non‐sufficient condition for an afternoon maximum of convection; limited spatial water vapor variability is also necessary. The imposition of the mean diurnal cycle has no statistically significant impact on the mean timing of clustering onset.


  • Open Access

    Impact of a mixed ocean layer and the diurnal cycle on convective clustering

    by Adrian Mark Tompkins, Addisu Gezahegn Semie

    We investigate how interactive surface feedbacks impact convective clustering in a cloud resolving model coupled to a slab ocean, where the domain-mean temperature is controlled with an adaptive Q-flux. In the first investigation, with constant domain-mean surface temperature, progressively thinner ocean layers slow the onset of clustering by up to a month through surface feedback. Enhanced solar radiation in nascent dry, cloud-free areas leads to local surface warming, increasing latent heat fluxes in regions distant from convection. Once the magnitude of humidity anomalies increases, longwave emission dominates and the ocean cools rapidly under the dry patches. In this stage the surface feedback reverses and favors clustering. In the second investigation using a 1 meter mixed layer, the ocean undergoes a diurnal cycle in response to solar forcing, with a diurnal range of 2.5$^o$C in the domain mean. This leads to convective rainfall shifting from a weak broad nocturnal maximum to a sharper afternoon peak. The consequent daytime anvil shielding reduces the spatial SST variance, but the sharper rainfall peak results from more concurrent convective towers, which are distributed throughout the domain due to coldpools. This reduces the upper tropospheric water vapor variance and causes a slight delay in clustering onset, despite the reduced spatial SST variance. We find that the onset time is deterministic when strong forcing causes clustering to occur quickly (after 25 days), whereas it is highly stochastic when surface feedback weakens the diabatic forcing and delays clustering past 40 days.