%0 Journal Article
%9 ACL : Articles dans des revues avec comité de lecture répertoriées par l'AERES
%A Ogilvie, Andrew
%A Belaud, G.
%A Massuel, Sylvain
%A Mulligan, M.
%A Le Goulven, Patrick
%A Calvez, Roger
%T Surface water monitoring in small water bodies : potential and limits of multi-sensor Landsat time series
%D 2018
%L fdi:010073763
%G ENG
%J Hydrology and Earth System Sciences
%@ 1027-5606
%K TUNISIE ; ZONE SEMIARIDE ; MERGUELLIL BASSIN VERSANT
%M ISI:000442118900001
%N 8
%P 4349-4380
%R 10.5194/hess-22-4349-2018
%U https://www.documentation.ird.fr/hor/fdi:010073763
%> https://horizon.documentation.ird.fr/exl-doc/pleins_textes/divers18-09/010073763.pdf
%V 22
%W Horizon (IRD)
%X Hydrometric monitoring of small water bodies (110 ha) remains rare, due to their limited size and large numbers, preventing accurate assessments of their agricultural potential or their cumulative influence in watershed hydrology. Landsat imagery has shown its potential to support mapping of small water bodies, but the influence of their limited surface areas, vegetation growth, and rapid flood dynamics on long-term surface water monitoring remains unquantified. A semi-automated method is developed here to assess and optimize the potential of multi-sensor Landsat time series to monitor surface water extent and mean water availability in these small water bodies. Extensive hydrometric field data (1999-2014) for seven small reservoirs within the Merguellil catchment in central Tunisia and SPOT imagery are used to calibrate the method and explore its limits. The Modified Normalised Difference Water Index (MNDWI) is shown out of six commonly used water detection indices to provide high overall accuracy and threshold stability during high and low floods, leading to a mean surface area error below 15 %. Applied to 546 Landsat 5, 7, and 8 images over 19992014, the method reproduces surface water extent variations across small lakes with high skill (R-2 = 0.9) and a mean root mean square error (RMSE) of 9300m(2). Comparison with published global water datasets reveals a mean RMSE of 21 800m(2) (+ 134 %) on the same lakes and highlights the value of a tailored MNDWI approach to improve hydrological monitoring in small lakes and reduce omission errors of flooded vegetation. The rise in relative errors due to the larger proportion and influence of mixed pixels restricts surface water monitoring below 3 ha with Landsat (Normalised RMSE = 27 %). Interferences from clouds and scan line corrector failure on ETM+ after 2003 also decrease the number of operational images by 51 %, reducing performance on lakes with rapid flood declines. Combining Landsat observations with 10m pansharpened Sentinel-2 imagery further reduces RMSE to 5200m(2), displaying the increased opportunities for surface water monitoring in small water bodies after 2015.
%$ 062 ; 126