Conventional gradient methods have already been applied to reservoir engineering for matching the history of former field performances. The key point of these methods is to select the best arcal reservoir zoning for reduction of the amount of reservoir parameters to be identified. In this paper we propose a zoning based on reservoir lithofacies, thus making a more natural than geographical choice. The gradients of the bottom hole well pressure with respect to the reservoir characteristics to be identified are required by all history matching processes. These gradients are calculated using either a perturbation method or a so-called analytical method. Without any reservoir zoning both methods require as many problem solving jobs as grid blocks. In this article we propose computing gradients relating to lithofacies. Starting with a geological model in lithofacies that is generally from a seismic interpretation and pixel or object-based geostatistical simulation we try to quantify the effect of a lithofacies on well pressure response or production history. The influence of a lithofacies is measured through its petrophysical parameters. This method drastically reduces the number of parameters to be identified and allows the use of the natural partition of the reservoir into geological bodies. The gradient calculation relating to lithofacies has been successfully implemented in an implicit single-phase fluid flow model. This model can be used for well-tests simulations. By introducing gradients to minimize an objective function that measures the difference between observed and simulated well pressure responses, wc can effectively achieve the inversion of petrophysical lithofacies parameters. Several examples of inversion are given at the end of the article to illustrate the effectiveness of this gradient method.