Spatial processing techniques for satellite altimetry applications in continental hydrology

摘要

Radar-based satellite altimetry is a well recognized measuring technique with good precision for oceanographic applications. For continental hydrology, its use is complicated by a number of factors such as river width, satellite crossing angle and noise from the river banks or islands. These factors make precision vary significantly. The satellite crossing points can be made into virtual gauging stations that can complement the existing network of in situ stations. This article describes a series of spatially explicit processing to correct or exclude altimetry measurements not related to the water level. While some processing take advantage of a priori information such as the center line of the river, other processing are based on pattern recognition to characterize the shape described by the sequence of points. These problems are dealt with by fitting a second degree polynomial curve to the sequence of points and characterizing its shape. The correction is applied by determining a weight for each point in the crossing sequence of measurements. These processing approaches have been combined into a single tool called VHSTOOL. The method is tested on a 1000 km stretch of the Sao Francisco River in Brazil. Data from Envisat cover the 2003-2010 period while the recently launched Altika sensor provided data for a few months in 2013. Results show that the average accuracy of 60 cm obtained (45 cm by removing outliers) is comparable to that of completely manual methods. Altika measurements could not be validated since no recent in situ data was available but initial evaluation suggests increased details should bring some improvements over Envisat data.