Densely populated arid urban areas continue to expand, increasing the competing demand for limited water resources. Persistent water scarcity requires effective management and conservation strategies. Crop Coefficients (Kc) contribute to quantifying the amount of water lost through evapotranspiration (ET) from specific plants. However, most current research on Kc focuses on controlled agricultural environments, which align well with traditional biophysical models. Conversely, urban environments present unique challenges due to the diverse landscape, urban development, and heterogeneous vegetation. In this work, we propose a novel Machine Learning (ML) approach to calculate urban Kc using two types of ET, ground weather stations data, and remote sensing data. Reference Evapotranspiration (ETo) models were created using the Random Forest Regressor (RFR) algorithm and the Multilayer Perceptron (MLP) model. The ETo models were trained using weather data collected from 26 ground weather stations. The RFR and MLP models achieved a test R2 of 0.9549 and 0.9544, as well as a validation R2 of 0.9649 and 0.9169, respectively. These models were applied to estimate ETo in the study area using remote sensing data as input. Although the study area is not directly covered by the ground weather stations, it shares similar climatic and geographic characteristics and lies within the same urban footprint. ETo estimates are then combined with Actual Evapotranspiration (ETa) derived from remote sensing data to estimate Kc values with a resolution of 30 m2 across the study area. This approach offers a scalable solution for estimating Kc in urban environments with limited ground data and can be adapted to other cities to assess vegetation dynamics and water requirements, particularly in regions facing water scarcity.