The article by Yang et al., (2003) makes use of existing data (mainly from USGS) and estimation methods suggested from literature to estimate soil erosion at the global level. The estimation is based on Revised Universal Soil Loss Equation (RUSLE). The researchers simulated several scenarios from the past (1900s - 1980s), present (1980s) to the future (2090s, 2x of CO2) (see Table 2 of the article):

1. Scenario H: simulation based on historical land use and climate.
2. Scenario C: simulation based on present land use and climate.
3. Scenario D: simulation based on present land use and future climatic condition.
4. Scenario P: simulation based on potential vegetation under present climate.
5. Scenario F: simulation based on future land use.
6. Scenario E: simulation based on the future land use and predicted climatic condition.
   

The work on simulating various scenarios echoes with Bronstert's (2004) call for more work on simulation modeling.

This paper provides clear account on how global estimations were generated. The paper, however, seems to be vague about the contribution to our understandings of the "forcing" and how various components are interrelated to contribute to what we are experiencing right now. USing Bronstert's framework, this meanings that future research is needed to examine feedback between components of the system. It is also interesting to examine how the results match with estimation at the regional scale and explore the cause of the potential mismatches.

References
Bronstert, A, 2004, Rainfall-runoff modelling for assessing impacts of climate and land-use change, Hydrological Processes, 18, 567-570

Yang, D., Kanae, S., Oki, T., Koike, T., and Musiake, K., 2003, Global potential soil erosion with reference to land use and climate changes, Hydrological Processes, 17(14), 2913-2928 [doi]