Title: (From AGU) Using continuous sap flux and soil respiration datasets to infer the strength and speed of root-soil coupling in a deciduous forest

Authors: Stephanie C. Pennington, Ben Bond-Lamberty, Charlotte Grossiord, Wenzhi Wang, Nate McDowell

Target Journal:

Overall Scientific Question: What is strength and speed of above-belowground coupling (Js to Rs)?

Data Prep and Preliminary Figures:

Diagnostic plots for weather data

Hysteresis between temperature and Rs / Js

Arrows indicate the direction of the loop

Js versus Rs

Split by daytime and nighttime values

Raw data over time

PACF

The partial autocorrelation function gives the partial correlation (i.e. after controlling for other variables, or in this case, time lags) of a stationary time series with its own lagged values.

Science Questions.

Q1: For the overall dataset, how correlated are Js and Rs, at what time lags?

(Between the fluxes; here and afterward, on a per-tree basis.)

  • Whole dataset
  • Compute lag correlation for each tree - each timestamp hour

H1.1. Hypothesis-time.

Js and Rs will be correlated at some lag of (probably) multiple hours, because of the time it takes for sap to ascend; photosynthesis to occur; phloem to descend to roots; respiration to occur; and resulting CO2 to diffuse to soil surface .

H1.2 Hypothesis-species.

We expect there to be differences in peak lag and correlation between the two species - Tulip Poplar and Red Maple - driven by path length difference and light availability.

Tree Species Hour Lag Maximum Correlation
C3 LITU 13 0.32
C6 LITU 21 0.18
C7 ACRU 15 0.32
C8 ACRU 17 0.11

Q2: How does this change over the course of the growing season?

  • For each week of year, calculate the correlation between Js and Rs for each hour lag (using function) and pull out max correlation lag
  • Dendrometer data to show growth changes?

H2.1. Hypothesis.

We expect to see changes in the strength and speed of coupling, probably because of seasonal changes in photosynthetic capacity and carbon allocation (e.g. reflected in stem diameter growth data).

  • note: this dataset includes data from 2018 (wet year) AND 2019 (avg year w drought in July), which might be why theres some stronger correlations. might be useful to split this graph by year as well

Q3: Is this correlation or causation?

H3.1. Hypothesis.

If H1.1 is correct, then days with more sunlight would have a stronger correlation.

To examine this issue, we look at matched days, i.e. that are in the same part of the growing season and have similar conditions EXCEPT for sunlight.

Goal: To compare Rs:Js correlation of sunny vs. cloudy days. We essentially are testing the importance of PAR on the relationship.

  • Rs, Js, and climate variables parsed to same timescale
  • Data separated into sunny days (top 1/3 daytime PAR) and cloudy days (bottom 1/3 daytime PAR)
  • Just the matched days, how do the max-cor-lags differ between them?
  • Tree, DOY, Match_doy, Max_Cor, Lag_max_cor, Coverage

Constraint sensitivity test

We’ve created a function similar_days to match days of similar climate conditions but varying PAR (i.e. sunny-cloudy days)

First, we tested how the constraints for climate conditions impacted the number of matches returned

Constraint max min sd Constraint value
Precip 1.375 0.000 0.198 NA
RH 98.509 19.829 16.466 10.0
SM 0.494 0.144 0.075 0.2
T5 26.039 0.617 7.954 2.0
Tair 32.382 -9.768 9.579 2.0
VPD 1.712 0.014 0.342 0.5
Lookahead NA NA NA 10.0

Sunny-cloudy comparison

Stats

Differences in daily means between sunny and cloudy days? We use a paired (because we have matched days) Student’s t-test for this.

## Js_avg:
## 
##  Paired t-test
## 
## data:  tt_js$Cloudy and tt_js$Sunny
## t = -2.5667, df = 26, p-value = 0.01638
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -3.117715e-06 -3.447635e-07
## sample estimates:
## mean of the differences 
##           -1.731239e-06
## Rs_avg:
## 
##  Paired t-test
## 
## data:  tt_js$Cloudy and tt_js$Sunny
## t = 2.5345, df = 43, p-value = 0.01498
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.1188971 1.0449492
## sample estimates:
## mean of the differences 
##               0.5819232

Breakdown of each matched day

Matched days by Tree (columns) and day of the year of the sunny day (rows). Sunny days are shown in blue with it’s cloudy matches in red.

Matched days by day of the year of the sunny day. Sunny days are shown in blue with it’s cloudy matches in red.

Sunny-cloudy Q10