We revisit the concept of scalar gravity ano- maly determination by an airborne strapdown INS - GNSS system. We built on the previously investigated concepts (mainly within 1995-2005 period) while trying to decrease the error spectrum of the system caused by accelerometer biases at lower frequencies and GNSS- position/velocity noise at shorter wavelengths. We claim that the determination of the random long-term ac- celerometer bias is possible through combination of - GRACE + GOCE data that provide an unbiased field with 80 km resolution while the decrease in velocity noise is expected by precise-point-positioning (PPP) method that merges satellite-phase observations from GPS and Galileo. In the absence of Galileo constellation we focus our practical demonstration on the gravity- anomaly determination via INS/GNSS data filtering. We present first the modeling of an extended Kalman filter/smoother that determines the gravity anomaly to- gether with the trajectory, which is a preferred method over the cascade determination (i.e. separate estima- tion of trajectory and specific forces, GNSS acceleration and low-pass filtering of the merged signal). Second, we show how to incorporate the same modeling within the concept of dynamic networks. This approach al- lows imposing cross-over conditions on the state of grav- ity anomaly at trajectory intersections while estimating the sensor and trajectory errors at the same time. This is indeed rigorous formulation of the problem that is expected to surpass the conditioning via cross-over ad- justment that in previous investigations followed the fil- tering/smoothing. Despite the remaining challenges of the method of dynamic network caused by large num- ber of parameters (i.e. > 106), we present first practical results obtained within European FP7 GAL project.