Rs_bahn predicted by TS, TAnnual, and MAT comparison

[jinshijian]

• From the monthly global soil respiration database (MGRsD), there are totally 325 records from 68 papers can get annual average soil temperature (measure month >= 1 yr).
• Based on those data, I compared Rs_bahn predicted based on Bahn 2010 equation, but using TS_Annual, TAnnual (annual air temperature from Delaware climate data, I did not use study_temp because we have only less than 10 measurements available), and MAT (from Delaware climate data), please see the table and figure.

Parameter	Intercept	P_inter	Intercept	P_inter	R2	N
Soil temperature	-1.130	0.961	1.084	0.0004	0.866	325
TAnnual_Del	35.046	0.434	0.946	0.2430	0.563	325
MAT_Del	36.54	0.278	0.909	0.0094	0.678	325

• Generally, using soil temperature got highest R^2, but slope significantly larger than 1.
• Using TAnnual has smallest R^2 but slope does not significantly differ from 1.
• Using MAT slope significantly smaller than 1, has a middle level of R^2 value.

• If move agriculture sites (105 records left), slope = 1.01 and intercept = 37.02, (P>0.05 for both).

• RH-dominated ones (RC_annual>0.5 is FALSE) significantly off the trend, why?

• When getting drought (SPI decrease), slope gets larger and larger (tend to overestimate).

To do?

From SRDB, there are 168 studies available (about 100 papers not available from MGRsD), and I checked those papers, and find that soil temperature data is available for most of those paper, should I digitize those paper and get the annual soil temperature data? (I think it works to do it, because later we want to separate the data into different subgroups, some groups do not have enough data. It should not take too much time, one week should be able to get all 100 papers data, what do you think?)

[jinshijian]

To do?

Another thing I am thinking is, for some sites only has less than 12 month soil temperature data (e.g., > 7 month's data), can we built a regression between Soil temperature and Air temperature, and based on the Air temperature to predict the missing months' soil temperature and then get a annual soil temperature?