2.3 Methods
1)Linear
regression model
We used linear trend analysis of the NDVI, NPP-A,
NPP-P, and NPP residuals time series to investigate
vegetation growth. The slope (a ) was used to express the change
trend, and the p value was used to express the significance level of the
change. The linear trend analysis was then extended to a grid scale. We
calculated the slope according to Dai et al. (2014) as follows:
\(a=\frac{\sum_{i=1}^{n}{(x_{i}-x)(y_{i}-y)}}{\sum_{i=1}^{n}{(x_{i}-x)}^{2}}\),
where a is the trend (slope) of the change, with a positive
(negative) value indicating an increasing (decreasing) vegetation trend.
Taking NDVI as an example, x is the average NDVI value, \(x_{i}\)is the regional average NDVI value in the ithyear, and y is the year. We classified the resultant slope values
into four categories: significant increase (a > 0,P < 0.05); non-significant increase (a> 0, P > 0.05); significant decrease
(a < 0, P < 0.05); non-significant
decrease (a < 0, P > 0.05).