Peatland Water Table Depth Monitoring Using Quad-Polarimetric L-Band SAR

Accurate, remote sensing-based methods of monitoring peatland water table depth (WTD) could support peatland management and restoration. Quad-polarimetric L-band synthetic-aperture radar (quad-pol L-band SAR) imagery has potential advantages over other forms of SAR more often used for WTD modeling. Using quad-pol L-band SAR observables as predictors, we modeled spatiotemporal and temporal WTD variation at a raised bog in Scotland, evaluating model performance, and several parameters' effects on performance. While predicting WTD spatiotemporally will likely remain challenging (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.02) until the effects of vegetation and other characteristics on SAR signal can be accounted for, we found strong ability (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.69) to predict temporal WTD variation that exceeded the performance found by several previous studies that used C-band SAR or optical imagery. This could enable wide-scale peatland monitoring identifying when areas have low or high WTDs relative to normal, and are more or less vulnerable to degradation, respectively. Model performance was validated using leave-location-out cross validation (LLOCV) rather than leave-one-out cross validation as the latter was found to inflate performance estimates. LLOCV uses spatial blocking to reduce the exaggeration in model performance estimates caused by spatial autocorrelation, and, due to the structure of many WTD datasets, we recommend future work in this area employs LLOCV or a similar technique. A validation technique that accounts for both spatial and temporal autocorrelation would further improve confidence in results.