Calibrating an evapotranspiration model using radiometric surface temperature, vegetation cover fraction and near-surface soil moisture data

2018

Article référencé dans le Web of Science WOS:000437060700011

Hssaine B. A., Merlin Olivier, Rafi Z., Ezzahar J., Jarlan Lionel, Khabba S., Er-Raki S.

Agricultural and Forest Meteorology, 2018, 256, p. 104-115 ISSN 0168-1923

An accurate representation of the partitioning between soil evaporation and plant transpiration is an asset for modeling crop evapotranspiration (ET) along the agricultural season. The Two-Surface energy Balance (TSEB) model operates the ET partitioning by using the land surface temperature (LST), vegetation cover fraction (fc), and the Priestley Taylor (PT) assumption that relates transpiration to net radiation via a fixed PT coefficient (alpha(pT)). To help constrain the evaporation/transpiration partition of TSEB, a new model (named TSEB-SM) is developed by using, in addition to LST and fc data, the near-surface soil moisture (SM) as an extra constraint on soil evaporation. An innovative calibration procedure is proposed to retrieve three key parameters: an and the parameters (a(rss) and b(rss)) of a soil resistance formulation. Specifically, a(rss) and b(rss) are retrieved at the seasonal time scale from SM and LST data with f(c) < 0.5, while apT is retrieved at the daily time scale from SM and LST data for f, > 0.5. The new ET model named TSEB-SM is tested over 1 flood- and 2 drip-irrigated wheat fields using in situ data collected during two field experiments in 2002-2003 and 2016-2017. The calibration algorithm is found to be remarkably stable as an, a and brss parameters converge rapidly in few (2-3) iterations. Retrieved values of alpha(pT), a(rss) and b(rss) are in the range 0.0-1.4, 5.7-9.5, and 1.4-6.9, respectively. Calibrated daily apT mainly follows the phenology of winter wheat crop with a maximum value coincident with the full development of green biomass and a minimum value reached at harvest. The temporal variations of alpha(pT) before senescence are attributed to the dynamics of both root-zone soil moisture. Moreover, the overall (for the three sites) root mean square difference between the ET simulated by TSEB-SM and eddy-covariance measurements is 67 W M-2 (24% relative error), compared to 108 W m(-2) (38% relative error) for the original version of TSEB using default parameterization (alpha(pT) = 1.26). Such a calibration strategy has great potential for applications at multiple scales using remote sensing data including thermal-derived LST, solar reflectance-derived f(c) and microwave-derived SM.

Sciences fondamentales / Techniques d'analyse et de recherche [020] ; Bioclimatologie [072] ; Sciences du monde végétal [076] ; Télédétection [126]

MAROC

Fonds IRD [F B010073217]

fdi:010073217