Weather Forecasts Using Deep Learning Models for Urban Areas in Nigeria
Domaine:
climateType de record:
paper
Créateur:
TayOpe
Éditeur:
Int
Hôte:
The data includes information from ten years, from 2014 up to 2023. Weather data is recorded monthly from the
meteorological station at the Forestry Research Institute in Nigeria. Important data recorded includes temperature in
Celsius, humidity, wind speed in km/h, and relative humidity. This data is recorded in seven columns and 120 rows. It
includes temperature, humidity, wind speed, and relative humidity. Three models are applied in solving this problem. These
models are time-aware long short-term memory networks, gated recurrent units, and the Transformer. These models have
been improved with an attention-based approach to interpretability, inspired by RETAIN. The GRU model can forecast
data up to six months into the future. This data shows an inverse correlation with temperature and relative humidity.
Relative humidity goes down to 72% ± 5, indicating pre-rainy conditions. This occurs while temperatures peak at 31°C ± 0.8
in Month 4. In Month 6, temperatures drop to 28.2°C ± 0.5, and relative humidity rises to 85% ± 3, indicating that rain is
on the way. Wind speeds decrease to 9.8-10.5 km/h in Months 3 and 4, when temperatures are at their peak and relative
humidity is at its lowest. The weather forecasting model has shown how GRU, Time Aware, and even the Transformer can
be applied in solving weather problems in Nigeria.
Similaires
Classifying Economic Areas for Urban Planning using Deep Learning and Satellite Imagery in East AfricaHazardous Atmospheric Dispersion in Urban Areas: a Deep Learning Approach for Emergency Pollution ForecastPLANT DISEASE CLASSIFICATION USING DEEP LEARNING MODELSMaize Disease Detection using Deep Learning ModelsLocalized Weather Prediction Using Kolmogorov-Arnold Network-Based Models and Deep RNNsDeep Learning improved seasonal forecasts for the Blue Nile Basin