In this chapter we focus on the observations that are available for operational, real-time applications in meteorology, i.e., for numerical weather prediction (NWP). Many in situ observations can be treated as point-wise measurements. Their influence on the analysis is expected to be localized and smoothed according to the specified background error covariance structures (chapters Mathematical Concepts of Data Assimilation, Nichols; Error Statistics in Data Assimilation: Estimation and Modelling, Buehner). Most remotely-sensed sounding data, on the other hand, are integrated measurements that cannot be treated as point-wise observations. This is an important distinction which needs to be accounted for by the analysis scheme. Therefore, we expand the discussion of observation operators to integrals, and examine how such data can be expected to influence the analysis. In operational meteorology, the most prominent examples of integral observations are measurements of infrared and microwave radiation from satellite instruments. Other, recent examples include ground-based GPS (Global Positioning Satellites) and radio-occultation data. The related issues of quality control and data thinning are also covered. Assimilation of time-sequences of observations is discussed. This chapter complements chapters The Global Observing System (Thepaut and Andersson) and Research Satellites (Lahoz).