MODIS land cover uncertainty in regional climate simulations

• Published in: Climate dynamics Vol. 49; no. 11-12; pp. 4047 - 4059
• Main Authors: Li, Xue, Messina, Joseph P., Moore, Nathan J., Fan, Peilei, Shortridge, Ashton M.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01-12-2017; Springer; Springer Nature B.V
• Subjects: Agricultural land; Analysis; Atmospheric models; Barren lands; Climate; Climate change; Climate models; Climatology; Computer simulation; Earth and Environmental Science; Earth Sciences; Geographical variations; Geophysics/Geodesy; Grasslands; Heat flux; Heat transfer; Land cover; Latent heat; Latent heat flux; Modelling; MODIS; Oceanography; Rams; Regional climates; Simulation; Soil; Soil moisture; Trajectory analysis; Uncertainty; Uncertainty analysis; China; Uncertainty propagation; Regional climate simulation; RAMS; Land cover; MCD12Q1
• ISSN: 0930-7575; 1432-0894
• DOI: 10.1007/s00382-017-3563-7

MODIS land cover datasets are used extensively across the climate modeling community, but inherent uncertainties and associated propagating impacts are rarely discussed. This paper modeled uncertainties embedded within the annual MODIS Land Cover Type (MCD12Q1) products and propagated these uncertainties through the Regional Atmospheric Modeling System (RAMS). First, land cover uncertainties were modeled using pixel-based trajectory analyses from a time series of MCD12Q1 for Urumqi, China. Second, alternative land cover maps were produced based on these categorical uncertainties and passed into RAMS. Finally, simulations from RAMS were analyzed temporally and spatially to reveal impacts. Our study found that MCD12Q1 struggles to discriminate between grasslands and croplands or grasslands and barren in this study area. Such categorical uncertainties have significant impacts on regional climate model outputs. All climate variables examined demonstrated impact across the various regions, with latent heat flux affected most with a magnitude of 4.32 W/m
2
in domain average. Impacted areas were spatially connected to locations of greater land cover uncertainty. Both biophysical characteristics and soil moisture settings in regard to land cover types contribute to the variations among simulations. These results indicate that formal land cover uncertainty analysis should be included in MCD12Q1-fed climate modeling as a routine procedure.