Dedan Kimathi University of Technology

Gully erosion is a major driver of land degradation and sediment flux in transitional highland environments, yet its magnitude and time dynamics are poorly understood in East Africa. This study quantified multi-temporal gully expansion and volumetric soil loss in the Kakia and Esamburmbur sub-catchments (Enkare-Narok Basin, Kenya) using an integrated Earth Observation framework adopting very-high-resolution imagery (2009-2024), UAV photogrammetry (0.10 m DEM), and DEM-of-Difference (DoD) analysis. The mapped gully area increased from 5.93 ha in 2009 to 14.94 ha in 2024, with the largest mapped extent occurring in 2019 (16.60 ha). UAV-derived morphometry revealed substantial geometric variability, with a mean depth of 1.02 m and a maximum width of 40 m. DoD analysis estimated erosion volumes of 116,633 m3 (FABDEM) and 130,497 m3 (WorldDEM) within the UAV survey footprint. High-volume gully identification flagged 39 gullies exceeding a threshold of 1665.20 m3 and accounting for a combined predicted erosion volume of 203,345 m3. A near-linear area-volume scaling relationship (b = 1.02, R2 = 0.41) enabled extrapolation of volumetric estimates beyond the UAV footprint. Scaling-derived estimates indicated increasing erosion volumes between 2009 and 2019, followed by a reduction by 2024. Comparison of DEM sources showed an uncertainty range of approximately 11.9-13.3% in volumetric estimates, while RTK-GNSS validation confirmed high vertical accuracy of the UAV-derived DEM (RMSE = 0.062 m). The results show the value of integrating multi-temporal imagery with UAV-derived topography to link longterm gully expansion with volumetric erosion assessment. Although uncertainties remain regarding DEM-source selection and scaling hypotheses, the framework provides a practical approach for evaluating erosion magnitude and spatial concentration in data-limited environments.

Solar-powered irrigation Systems (SPIS) are critical for agricultural production enhancement, food security and climate change adaptation, especially in Arid and Semi-Arid Lands (ASAL). There is increased attention towards shifting to more abundant and cleaner energy potential sources for revitalising irrigation strategies in ASAL areas. This study employed a geospatial approach to identify suitable locations for solar-powered irrigation systems (SPIS) in Baringo County, Kenya. Based on an integrated use of GIS spatial analysis and analytical hierarchy procedure (AHP), suitable locations for solar-powered irrigation were mapped. Precipitation, irrigated areas, proximity to rivers, slope, and solar radiation were analysed and processed to derive spatially explicit SPIS suitability classes ranging from very low to very high suitability. The thematic layers were assigned weights based on Saaty’s AHP method, where weights for each factor were determined from a pairwise comparison matrix, and a Weighted Linear Combination (WLC) approach was used to derive the final suitability classes for the county. The findings reveal that approximately 58

Accurate and explainable household load forecasting is critical for demand-side management, tariff-aware scheduling, and reliable smart grid operation. This study introduces a leakage-controlled multi-horizon forecasting pipeline that integrates predictive accuracy with statistical validation, interpretability, robustness, and operational relevance. We model multivariate household demand using an hourly smart-meter dataset spanning 14 months (Nov 2022-Jan 2024; N=10,234 time steps), incorporating aligned local weather covariates. A hybrid Transformer-BiLSTM is trained using a multi-output configuration to predict 24-hour and 168-hour trajectories. Hyperparameters are optimized via Bayesian optimization (Optuna) employing chronological train/validation/test splits and a rolling-origin evaluation protocol. Performance is assessed using MAE, RMSE, and MAPE, while pairwise forecast differences are validated using the Diebold–Mariano procedure. Model explanations are generated through SHAP and attention analyses, further complemented by robustness testing (noise and feature dropout) and inference-efficiency measurements. At the 24-hour horizon, the hybrid model achieves MAE and RMSE values of 0.0539/0.0701 (MAPE 29.7

AfriVoices-KE is a large-scale multilingual speech dataset comprising approximately 3,000 hours of audio across five Kenyan languages: Dholuo, Kikuyu, Kalenjin, Maasai, and Somali. The dataset includes 750 hours of scripted speech and 2,250 hours of spontaneous speech, collected from 4,777 native speakers across diverse regions and demographics. This work addresses the critical underrepresentation of African languages in speech technology by providing a high-quality, linguistically diverse resource. Data collection followed a dual methodology: scripted recordings drew from compiled text corpora, translations, and domain-specific generated sentences spanning eleven domains relevant to the Kenyan context, while unscripted speech was elicited through textual and image prompts to capture natural linguistic variation and dialectal nuances. A customized mobile application enabled contributors to record using smartphones. Quality assurance operated at multiple layers, encompassing automated signal-to-noise ratio validation prior to recording and human review for content accuracy. Though the project encountered challenges common to low-resource settings, including unreliable infrastructure, device compatibility issues, and community trust barriers, these were mitigated through local mobilizers, stakeholder partnerships, and adaptive training protocols. AfriVoices-KE provides a foundational resource for developing inclusive automatic speech recognition and text-to-speech systems, while advancing the digital preservation of Kenya's linguistic heritage.

This research presents a behaviorally informed framework for synthesizing financial time-series data, specifically designed to emulate the complex dynamics of foreign exchange markets. Deviating from conventional generative adversarial networks (GANs) or purely statistical distribution-matching, the proposed methodology adopts a game-theoretic architecture. This framework integrates trader-interaction dynamics, stochastic strategies, and information asymmetry, treating the market as a strategic participant to reproduce authentic volatility patterns and structural dependencies. To ensure numerical stability across extensive simulations, the study introduces a uniform upward scaling procedure and controlled initialization, preventing pathological price behaviors without compromising the underlying statistical properties. The framework's analytical fidelity was rigorously evaluated against a suite of econometric and machine learning models, including ARIMA, XGBoost, LSTM, N-BEATS, and DLinear. Experimental results involving 12,960 hourly observations demonstrate that the synthetic data maintains strong alignment with empirical benchmarks. DLinear emerged as the superior model, exhibiting exceptional stability with an R2 frequently exceeding 0.98 and a Mean Absolute Scaled Error (MASE) near unity. While XGBoost and N-BEATS yielded competitive results, ARIMA and LSTM showed anticipated performance degradation due to temporal noise. Comprehensive residual diagnostics, including Ljung-Box tests and stationarity assessments, confirm that the generated series are behaviorally consistent and analytically reliable. This framework thus provides a robust foundation for comparative modeling and experimental financial research.