Abstract In this work we will explore the theoretical and practical aspects of nonparametric exponential deconvolution in the two-dimensional setting. After a transformation, exponential deconvolution can be used to estimate a decreasing density from direct observations. First, we will rigorously derive an inversion formula that can be implemented in an actual software implementation. We will then proceed with summarizing some of the known statistical properties of nonparametric kernel density estimators. Finally, we will combine the obtained results and construct, implement and test an exponential deconvolution method based on kernel estimators. An interesting application for estimation of decreasing densities will be pointed out and elaborated. Keywords: deconvolution, decreasing densities, kernel estimation.