4.1 Errors in the residual method
The vegetation change trend, evaluated with respect to the driving
factors and NDVI indicator, showed restoration in the MP and each in
subregion, but there were some differences in the restoration ratios
between those two evaluation methods. This section focuses on this
phenomenon, which can partly explain the “Errors” listed in Tables 1
and 3.
We classified possible scenarios as A, B, C, and D (see Table 1).
Scenario A occurs when the negative effect of climate change is greater
than the positive effect of human activity, resulting in vegetation
degradation. Scenario C occurs when the negative effect of human
activity is less than the positive effect of climate change, resulting
in vegetation recovery.
Results yielding scenarios B and D were defined as “Errors” here. NDVI
and actual NPP are both indexes that can reveal and track vegetation
growth and dynamics, and these two indexes are closely correlated. For
example, the research of De et al. (2013) captured the spatial pattern
of NPP, and the global trend of NPP is consistent with the NDVI for the
same period of time.
Rashid et al. (2016) revealed that about 68% of the global land area
has positive NPP values with an increasing trend, which corresponds
closely to the 67% of the MP land area with positive NDVI values and an
increasing trend. This phenomenon provides theoretical support for our
use of NDVI and NPP in the RESTREND method to validate the driving
forces of vegetation dynamics. However, there are inconsistencies
between NDVI and NPP values in some regions and situations. For example,
if grassland or forest is transformed into farmland/plantations, the
actual NPP could show a huge increase because of artificial
fertilization and irrigation. Consequently, it is possible that changes
in NDVI < 0 and changes in NPP-RES> 0 could co-occur if NPP-P increases but
NPP-A increases more, and vice versa. We believe that
this can explain part of the error scenarios outlined above from a data
perspective.