Any science must make predictions (that ultimately must measure true), whether climate science or soil science.  The part that makes this study a farce is the lack of ability to accurately predict decades of stover residue removal rate studies.   In his study, 1 -1=-1 ((since what he actually modeled was 1-2=-1))  see figure 1, where now dozens of publications show how and when 1-1=0 at practical, responsible removal rates in no till/strip till systems.    To my knowledge, every commercial crop residue operation sets strict agronomic practice guidelines (for nothing else that to save their own skin should non-point pollution come to litigation). 

I wouldn't go so far as calling all modeling disingenuous, but  a model output does NOT = data and it MUST predict a future state.  In the lack of ability to fast forward into the future, models are typically subjected to "leave one out" validation which the study makes not attempt to do, namely because there is no actual data set available to cross validate.   The author only goes so far as to validate SOC decomposition rates (which of course should approximate since it's is exactly what the model was built on, ie self referencing. 

The particular study is GIGO, but could be fixed to reflect accepted findings and best practices ALREADY in place.   There is good science out there, but reader , reviewer, and editor beware.   

My real beef is the reason his was published was to gain citations (half the reason Nature publishes the way it does is to keep it's impact factor) , career moves, , press exposure, maybe to stimulate discussion.  Heck, i'm doing it right now.    80% of the topics in high impact journals are very good science - the remaining 20% seem to be purely hit generators.  This is an age-old problem.