vapotranspiration estimation based on hourly LST observations acquired by a constellation of small satellites: a mission concept

FRANCISCO GRINGS; DARA ENTEKHABI ; ESTEBAN ROITBERG; PABLO PERNA; VERONICA BARRAZA; MARIANO FRANCO

Buenos Aires

Congreso; 1st IAA Latin American Symposium on Small Satellites; 2017

UNSAM, CONAE

Evapotranspiration (ET) is one of the most important processes representingecosystem-atmosphere interactions, a key element of the water cycle andland-surface energy balance. Its equivalent in energy is the latent heat flux(LE [W/m2]), which is the heat required to drive the total water vapour fluxfrom soil evaporation and plant transpiration. However, LE cannot bemeasured directly by remote sensing systems because it depends upon theavailability of water inside the and vegetation and energy to drive the waterstate change (from liquid to vapour). Therefore, at regional scales, landsurface models (LSM) or empirical relations driven by satellite derivedindices are used to estimate LE. Among them, spaceborne measurements ofland surface temperature (LST) provide an important constraint for theestimation of surface energy balance components over large spatialdomains.Since turbulent fluxes are components of the surface energy balance (SEB),the state variable of that balance (LST) is often the chosen pathway to mapthese fluxes based on remotely sensed sensor measurements. In order toestimate the components of the SEB from LST measurements, the SEB isitself often used as a constraint (for example, using the heat diffusionequation). The most straightforward approach to use the model as aconstraint of the estimation is to implement a variational data assimilation(VDA) scheme, which allows to combine dynamic SEB models with thesuccessive measurements of LST to estimate LE. In this context, the VDAscheme with SEB as the physical constraint and satellite LST estimates asobservations essentially finds the states and parameters that constitute amodel trajectory that most closely tracks the observations. The variationscheme runs trying to minimize the quadratic of model-observationmismatches. However, in weak-constraint VDA, the model and theobservations are taken as uncertain, in order to avoid overfitting. Thisapproach has distinct advantages for dealing with occasional erroneousforcing data (e.g., micrometeorological measurements) and structural errorsin the physical constraint (the suitability of using the heat diffusion equationto model the transfer of heat inside the vegetation canopy and soil).One important aspect of this technique it that it is based on the modeling ofthe hourly variation of LST in the local sunny hours (10 - 16 hs local time).Therefore, it requires hourly observations of LST. Therefore, the datarequired to implement this scheme cannot be provided by an instrumentmounted on a single heliosynchronous platform, since only two observationsper day are available (and one unusable (at night)). Therefore, the datarequired for this approach are usually obtained from instruments mountedon geosynchronous platform (i.e. GOES), which are generally expensive andrequire very high resolution IR cameras to provide a reasonable resolutionon the ground.In this framework, it is interesting and promising to investigate the potentialof small satellitesto obtain LST hourly information in a way to increase our ability to estimatedaily LE. This work propose the use of a constellation of 7 small satellites ona single orbit with a time offset of 1 hour (1 per hour from 10 hs to 16 hs). Inthis way, the hourly variation of LST can be observed at both high temporaland spatial resolution using relatively inexpensive cameras and platforms.The main assumption is that by taking advantages of small satellites hourlyLST observations, the estimation of LE could be improved. In this work wepresent simulations to test this hypothesis and its robustness based on theFirst International Satellite Land Surface Climatology (ISLSCP) FieldExperiment (FIFE), which is centered on a 15 x 15 km test site nearManhattan, Kansas.