The Performance and Potentials of the CryoSat-2 SAR and SARIn Modes for Lake Level Estimation

2017 Water  
Over the last few decades, satellite altimetry has proven to be valuable for monitoring lake levels. With the new generation of altimetry missions, CryoSat-2 and Sentinel-3, which operate in Synthetic Aperture Radar (SAR) and SAR Interferometric (SARIn) modes, the footprint size is reduced to approximately 300 m in the along-track direction. Here, the performance of these new modes is investigated in terms of uncertainty of the estimated water level from CryoSat-2 data and the agreement with in
more » ... situ data. The data quality is compared to conventional low resolution mode (LRM) altimetry products from Envisat, and the performance as a function of the lake area is tested. Based on a sample of 145 lakes with areas ranging from a few to several thousand km 2 , the CryoSat-2 results show an overall superior performance. For lakes with an area below 100 km 2 , the uncertainty of the lake levels is only half of that of the Envisat results. Generally, the CryoSat-2 lake levels also show a better agreement with the in situ data. The lower uncertainty of the CryoSat-2 results entails a more detailed description of water level variations. Water 2017, 9, 374 2 of 13 effect. Ref. [8] derived water level heights for both rivers and wetlands from TOPEX/Poseidon, and Ref. [9] used 10 Hz data from TOPEX/Poseidon to study water level changes over Louisiana vegetated wetlands between 1992 and 2002. Ref. [10] studied seasonal water level variability of boreal wetlands in Western Siberia from Envisat. Over time, the data quality and the methodology to process the data have greatly improved. Currently, root mean square error (RMSE) estimates of just a few cm are obtained for selected lakes when comparing with in situ data [11] . CryoSat-2 and the recently launched Sentinel-3 represent a new generation of altimetry missions. These satellites apply Synthetic Aperture Radar (SAR) technology [12] , which entails a reduction of the footprint in the along-track direction to approximately 300 m [13] . The smaller footprint size allows for monitoring much smaller lakes more accurately than previously. CryoSat-2 covers the Earth up to 88 degree latitude and has a repeat period of 369 days. The number of satellite crossings over a given lake therefore depends on the lake extent in the east-west direction and the latitude [14] . Hence, smaller lakes are not visited sufficiently to capture the seasonal signal. On the other hand, significantly more lakes are visited. Recently, some studies regarding lake level estimation including new processing strategies of CryoSat-2 data have been carried out. Ref. [15] presented a new waveform retracker based on cross-correlation of a modeled CryoSat-2 waveform with the observed waveforms. Ref. [16] demonstrated that the SAR mode provides an increased precision for small lakes compared to conventional altimetry. Ref. [11] presented a novel SAR mode retracker, which utilizes information from several waveforms simultaneously, and [17] demonstrated that waveform classification might be a powerful tool to handle erroneous data. Ref. [14, 18] used CryoSat-2 data to investigate the trend and seasonal signal of lakes on the Tibetan Plateau. Here, we intend to quantify the quality of CryoSat-2 data in the SAR and SARIn modes for lake level estimation and prove its better performance over smaller lakes compared to conventional altimetry from Envisat. This has previously only been done in studies where a few lakes were investigated [16, 17] . To quantify the quality of the lake levels derived from CryoSat-2, we perform a thorough investigation of the performance of CryoSat-2 compared to conventional altimetry as observed by Envisat. The study is based on a set of 145 lakes which are covered by both CryoSat-2 (SAR or SARIn mode) and Envisat (LRM). The lakes are located in Canada, Finland, and Denmark and have areas ranging from a few to several thousand km 2 . A way to evaluate the data is to consider the standard deviation of the predicted water level for each crossing over a given lake. For each lake, the standard deviations are summarized by the median, which hereafter is referred to as the median of standard deviation (MSD). The MSD gives a measure of how accurately the water level is estimated, which subsequently determines how small water level variations that can be observed. We estimate the MSD for each lake and test its dependence on lake area, in order to evaluate the improvement available with the new altimetry modes. In situ data is available for selected Canadian lakes, which enables the evaluation of the ability to capture annual and interannual signals. Finally, the mean water level of Danish lakes is evaluated against accurate laser scanner data.
doi:10.3390/w9060374 fatcat:q4zzbklqdjekjc5jyigbp3b4hu