The temporal changes of soil moisture at one subsurface layer where ground water is deep can be controlled by precipitation and plant transpiration. Here I want to quantify the contributions of daily precipitaiton (P), cumulative daily precipitation (CP), daily transpiration (T) and cumulative daily transpiration (CT) to soil moisture variations at 100 cm (θ). I would like to do this by using multivariate linear regression in SPSS. However, I find that the input variables of CP and CT are highly linear correlated. To my knowledge of statistics, one of them should be removed from inputs. But they should be retained according to physical process of water flux; they represent downward and upward water flux of controling soil water dynamics, respectively. Furthermore, I check that P is not linear correlated with θ (P>0.1). What can I do for these problems? Thanks a lot for any help in advance.
Since you are working with daily data, can you remove the cumulative values, and add in a value for incipient soil moisture (ie, calculated soil moisture value at day end). You may need to add in a factor to remove any excess or gravity water that is not retained long enough to be transpired. You did not mention if measuring transpiration or estimating based on a model.
Stepdown partial regression analysis will be more effective to screen the important variables contributing towards variation in soil moisture, to be classified on the basis of their contribution , either in increasing or decreasing order...
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS