Remote sensing can be very useful for spatialising a one dimension dynamic crop model at the scale of a territory. Spatialising a crop model can be done through a spatial estimation of its inputs. The comparison through time series of remote sensing observations and model simulations can be a way of spatialising model inputs by minimising the differences between observations and simulations (optimisation of inputs). It is strongly advisable to perform a sensitivity analysis to determine what are the inputs that have to be estimated This study is based on the sugarcane model MOSICAS and proposes a comparison between four methods of sensitivity analysis, which are quite easy to implement with crop models and well adapted to large scale simulations. The first one is a simple local method where the effect of input variation is studied independently. The three others are global methods. The second is based on a complete factorial design and on an ANOVA; the third one is a stochastic method and uses coefficients of correlation. The last one is also stochastic and calculates sensitivity index independently from a linear approximation of the model. The results show that the results of the different methods are consistent, i.e. all methods point out the same inputs, but on a more quantitative point of view, the order of sensitivity between methods may change. More over, the last two methods are more informative, but more time consuming too.