Despite the abundance and variety of remote sensing measurements, land surface characterization from satellite observations is still very challenging. The links between the three sources of surface information, namely the satellite observations, the in situ measurements, and the land surface model outputs, are complex. Innovative techniques have to be developed to merge these information sources and optimize the use of satellite measurements for better surface products and more predictability. Concepts such as multi-instrument/multiparameter retrieval algorithms are discussed, as well as the synergetic use of satellite observations, model outputs, and in situ data. The need for careful satellite calibration is stressed, and the scaling problem is emphasized. Recent results are reviewed to indicate what the land surface remote sensing problems are and how they might be attacked. Two concrete applications are presented: an "all weather" retrieval of surface skin temperature from combined microwave and infrared observations and a soil moisture analysis from the merging of multisatellite observations and land surface model outputs. This paper is intended to stimulate debates and collaborations between the land surface modelers and the satellite remote sensing community for the design of the next generation of land surface products.