Estimation of net primary productivity using a process-based model in Gansu Province, Northwest China

The ecological structure in the arid and semi-arid region of Northwest China with forest, grassland, agriculture, Gobi, and desert, is complex, vulnerable, and unstable. It is a challenging and sustaining job to keep the ecological structure and improve its ecological function. Net primary productivity (NPP) modeling can help to improve the understanding of the ecosystem, and therefore, improve ecological efficiency. The boreal ecosystem productivity simulator (BEPS) model provides the possibility of NPP modeling in terrestrial ecosystem, but it has some limitations for application in arid and semi-arid regions. In this paper, we improve the BEPS model, in terms of its water cycle by adding the processes of infiltration and surface runoff, to be applicable in arid and semi-arid regions. We model the NPP of forest, grass, and crop in Gansu Province as an experimental area in Northwest China in 2003 using the improved BEPS model, parameterized with moderate resolution remote sensing imageries and meteorological data. The modeled NPP using improved BEPS agrees better with the ground measurements in Qilian Mountain than that with original BEPS, with a higher R2 of 0.746 and lower root mean square error (RMSE) of 46.53 gC m−2 compared to R2 of 0.662 and RMSE of 60.19 gC m−2 from original BEPS. The modeled NPP of three vegetation types using improved BEPS shows evident differences compared to that using original BEPS, with the highest difference ratio of 9.21 % in forest and the lowest value of 4.29 % in crop. The difference ratios between different vegetation types lie on the dependence on natural water sources. The modeled NPP in five geographic zones using improved BEPS is higher than those with original BEPS, with higher difference ratio in dry zones and lower value in wet zones.