Gross primary production-coupled evapotranspiration in the global arid and semi-arid regions based on the NIRv index

•The model is optimized for global GPP and ET estimation with enhanced accuracy.•NIRv improves GPP precision by coupling with ET through fAPAR.•ET simulations outperform foundational and remote sensing models.•The model shows higher correlation and lower RMSE for GPP estimation using NIRv. In arid and semi-arid regions, accurate estimates of global primary productivity (GPP) and evapotranspiration (ET) are critical for understanding and managing water and carbon cycling in these fragile ecosystems. In this study, an improved ET-photosynthesis model (PT-JPL-GPP) was used to optimize GPP and ET estimates in these ecosystems by introducing the near infrared reflectance index (NIRv). NIRv, an indicator of the light use efficiency of vegetation, was integrated into the PT-JPL model. Compared to the original PT-JPL and existing remote sensing models, this PT-JPL-GPP model displayed a higher correlation (R2 = 0.73) and lower BIAS (−19.57 %) for GPP estimation. ET estimates were also noticeably improved, the R2 increased by 0.03(SN-Dhr) to 0.16(US-SRC), and the Root Mean Square Error (RMSE) reduced by 0.57 mm/month (SN-Dhr) to 4.64 mm/month (US-SRC). Particularly at the GRA site, the R2 was increased from 0.63 to 0.74, and the RMSE and bias was decreased by 1.25 mm/month and 10.51 %, respectively. The PT-JPL-GPP model was comparable with GLEAM, VPM, MOD17, MOD16, and PML-V2 models. The PT-JPL-GPP model exhibits a lower root mean square error and higher correlation for estimating GPP, compared to the VPM, MOD17, and PML-V2 models. The PT-JPL-GPP model outperformed PT-JPL, MOD16 models for estimating ET, but was slightly poorer than GLEAM and PML-V2 models. Our results highlight the merits of NIRv for improving GPP and ET estimates.