A remote sensing-based two-leaf canopy conductance model: Global optimization and applications in modeling gross primary productivity and evapotranspiration of crops

Remote-sensing-based (RS-based) Jarvis-type canopy conductance (Gc) of crops were parameterized coarsely on a regional or global scale. Two primary issues remain: (i) C3 and C4 crops were parameterized with the same parameter set; (ii) optimum conductance (gsmax) values were set to be temporally constant. To address the above issues, a RS-based two-leaf Jarvis-type canopy conductance model, RESTJA-Gc, was optimized for C3 and C4 crops on a daily basis, using hourly and daily flux data. Tower carbon flux and water flux were used to calculate Gc for optimizing RESTJA-Gc; RS-retrieved NDVI (RS-NDVI) was incorporated to estimate the temporal dynamics of gsmax . Our results showed that C3 and C4 crops responded differently to environments and Gc estimated by RESTJA-Gc agrees well with tower-flux retrieved Gc (R2=0.69, RMSE=0.0017 m ? s-1 ). RESTJA- Gc was applied in estimating LE and GPP. Daily mean meteorological inputs and daily mean Gc were generally incorporated into a Penman-Monteith (PM) approach to modeling daily canopy transpiration; however, such implementation may give unreasonable estimates. In this study, an existing evapotranspiration model was improved, RS- WBPM2, for estimating daily LE, and a reversed BBL model, R-BBL, incorporating the result of RS-WBPM2 was used to estimate daily GPP. Excellent agreements between flux tower measurements and modelled LE (R2 =0.77, RMSE=22.20 W ? m-2 for daily LE and R2 =0.85, RMSE=16.33 W ? m-2 for 16-day LE) and GPP (R2=0.81, RMSE=2.46 W ? m-2 for daily GPP and R2=0.83, RMSE=2.22 W ? m-2 for 16-day GPP) were obtained. Out of 19 flux sites under investigating, only one failed to reproduce model results. This failure was likely caused by special soil layer. Temporally dynamic gsmax in terms of NDVI, gsmax (NDVI), outperformed the temporally invariant gsmax , especially for GPP estimates.