Abstract. Green walls, façade greenery, living walls – vertical building greening as part of urban green infrastructure are measures for climate sensitive urban design, for water management and microclimate regulation. Strategic integration of green walls into local water and energy cycles requires prediction of evapotranspiration, considering the individual design, plant species, and building characteristics. Available models address horizontal surfaces but disregard vertical particularities and urban conditions, e.g., reduced direct radiation, spatial patterns of radiation on the wall due to building orientation and shading obstacles, and very heterogeneous wind fields that are influenced by rough surfaces, canyons, and adjacent wind barriers. We present a verticalization model, ET0vert, for the reference crop evapotranspiration ET0 (FAO) based on a sensitivity analysis. It comprises the adaptation of solar radiation and wind to the individual situations in front of a wall or facade. The accuracies of the model predictions are evaluated for (i) remote climate station data (horizontal reference plane), (ii) interpolated climate data (both horizontal and vertical reference plane) and (iii) on-site measured climate data (vertical reference plane, both not height-adapted and height-adapted) as input. We validate the model with data for a one-month reference period (25/07/2014 – 29/08/2014) from a weighable lysimeter with Fallopia baldschuanica greening of a 12 m high wall in Berlin, Germany. Regarding individual meteorological input parameters, we found high relevance of both vapor pressure deficit (VPD) and solar radiation (RS) for the study area. Using VPD and RS, respectively, a linear model could explain 90 % and 85 % of daily ET0 variances. No such relationship could be detected for wind speed, but for maximum and minimum wind speed. Compared to remote horizontal input data, verticalization of input data (RS and wind) reduced overestimations of ET from about 90 % to 14 % and 27 % for the daily and hourly resolution, respectively. If onsite climate data is available, deviations are reduced to 9 % and 5 % for the daily and hourly resolution. Height-adaptation of input data resulted in further improvements of the prediction accuracies (1 % and 2 % deviation for hourly and daily resolution). We conclude that simply using remote horizontal climate data for calculating ET of green walls is not advisable. Instead, any input data, onsite measured or remote climate station data, should be verticalized and preferably height-adapted. The verticalized model predicts the hourly and daily evapotranspiration of green walls necessary for e.g., irrigation planning, building energy simulations or local climate modeling.