Biofilm and leaves
Methods: Part I. Biofilm data: In 1-3 below you’ll find Chlorophyll, P uptake, and P/Chlo plots followed by simple linear models with ‘Site’ as random effect. I am currently searching alternative ways to analyze this data (responses are not normally distr.,log transform doesn’t improve that much and design is *not balanced anymore), but I would expect the trends hold if I used models that fit the data better. Part II. Leaf Litter data. Same issues apply.
Results: Part I. No direct effects of biofilm accrual were detetcted between open and close reaches but chlorophyll varied within reaches through a significant Stream Type *Canopy interaction (see 1, Chlorophyll). However, no significant effects on phosphorous uptake were detected, at least with a linear model. Part II. Leaves rom agricultural streams were significantly more fragmented than those from forested streams. Leaflitter decomposition followed a similar pattern, with more leaf mass remaining in leaves from forest than agricultural streams.
Part I. Biofilm accrual
1- Chlorophyll a
Quick Means (not transformed)
## Canopy Type N Chlorophyll sd se ci
## 1 Closed Agro 15 5.384880 5.8276220 1.5046855 3.2272295
## 2 Closed Forest 9 1.530978 1.1524944 0.3841648 0.8858857
## 3 Open Agro 15 17.972240 23.0947868 5.9630483 12.7894666
## 4 Open Forest 8 1.465950 0.7173697 0.2536285 0.5997361
Chlorophyll linear model
## Linear mixed model fit by REML t-tests use Satterthwaite approximations
## to degrees of freedom [lmerMod]
## Formula: LogChlorophyll ~ Type * Canopy + (1 | Name)
## Data: Balls
##
## REML criterion at convergence: 91.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.28908 -0.66304 -0.01105 0.56914 1.92414
##
## Random effects:
## Groups Name Variance Std.Dev.
## Name (Intercept) 0.9820 0.9910
## Residual 0.2671 0.5168
## Number of obs: 47, groups: Name, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.97871 0.45909 3.01000 2.132 0.12237
## Type1 0.73039 0.45909 3.01000 1.591 0.20943
## Canopy1 -0.22938 0.07861 40.01000 -2.918 0.00576 **
## Type1:Canopy1 -0.18368 0.07861 40.01000 -2.337 0.02456 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Type1 Canpy1
## Type1 -0.202
## Canopy1 -0.006 0.006
## Type1:Cnpy1 0.006 -0.006 -0.280Two-way ANOVA
## Analysis of Variance Table of type III with Satterthwaite
## approximation for degrees of freedom
## Sum Sq Mean Sq NumDF DenDF F.value Pr(>F)
## Type 0.67613 0.67613 1 3.014 2.5311 0.209428
## Canopy 2.27430 2.27430 1 40.011 8.5140 0.005757 **
## Type:Canopy 1.45832 1.45832 1 40.011 5.4593 0.024558 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Residuals from this basic Chlorophyll Model appear fine
=> Note the significant Canopy x Type interaction: reduced cholorphyll in open forests compared to open agro sites (? check)
2- Phosphorous uptake
Quick means (raw)
## Canopy Type N UP sd se ci
## 1 Closed Agro 15 1.276301 0.4389530 0.11333717 0.2430840
## 2 Closed Forest 9 1.210700 0.4239154 0.14130512 0.3258502
## 3 Open Agro 15 1.331308 0.3769649 0.09733192 0.2087562
## 4 Open Forest 9 1.406729 0.3365328 0.11217759 0.2586820
P uptake linear model
## Linear mixed model fit by REML t-tests use Satterthwaite approximations
## to degrees of freedom [lmerMod]
## Formula: UP ~ Type * Canopy + (1 | Name)
## Data: Balls
##
## REML criterion at convergence: 59.1
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.6207 -0.6461 -0.1466 0.6469 2.3837
##
## Random effects:
## Groups Name Variance Std.Dev.
## Name (Intercept) 0.009995 0.09998
## Residual 0.153200 0.39141
## Number of obs: 48, groups: Name, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 1.306259 0.074073 3.140000 17.635 0.000304 ***
## Type1 -0.002455 0.074073 3.140000 -0.033 0.975557
## Canopy1 -0.060715 0.058431 41.410000 -1.039 0.304792
## Type1:Canopy1 0.033212 0.058431 41.410000 0.568 0.572839
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Type1 Canpy1
## Type1 -0.231
## Canopy1 0.000 0.000
## Type1:Cnpy1 0.000 0.000 -0.252Two-way ANOVA
## Analysis of Variance Table of type III with Satterthwaite
## approximation for degrees of freedom
## Sum Sq Mean Sq NumDF DenDF F.value Pr(>F)
## Type 0.000168 0.000168 1 3.137 0.00110 0.9756
## Canopy 0.165411 0.165411 1 41.409 1.07971 0.3048
## Type:Canopy 0.049494 0.049494 1 41.409 0.32307 0.5728
=> No sign.differences in P uptake between reaches or land use types.
3- P Uptake per mg of Chlorophyll
Quick means (raw)
## Canopy Type N UPperClo sd se ci
## 1 Closed Agro 15 491.8667 449.8397 116.14811 249.1129
## 2 Closed Forest 9 1151.7723 664.3993 221.46642 510.7025
## 3 Open Agro 15 310.9156 332.3337 85.80818 184.0402
## 4 Open Forest 8 1145.5870 559.8127 197.92370 468.0152
P uptake by mg Chlorophyll linear model
## Linear mixed model fit by REML t-tests use Satterthwaite approximations
## to degrees of freedom [lmerMod]
## Formula: UPperClo ~ Type * Canopy + (1 | Name)
## Data: Balls
##
## REML criterion at convergence: 653.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.4401 -0.4485 -0.0893 0.2626 3.7867
##
## Random effects:
## Groups Name Variance Std.Dev.
## Name (Intercept) 146371 382.6
## Residual 138430 372.1
## Number of obs: 47, groups: Name, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 767.71 183.56 3.02 4.182 0.0245 *
## Type1 -366.32 183.56 3.02 -1.996 0.1392
## Canopy1 39.46 56.59 38.72 0.697 0.4898
## Type1:Canopy1 51.02 56.59 38.72 0.902 0.3729
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Type1 Canpy1
## Type1 -0.208
## Canopy1 -0.011 0.011
## Type1:Cnpy1 0.011 -0.011 -0.279Two-way Anova
## Analysis of Variance Table of type III with Satterthwaite
## approximation for degrees of freedom
## Sum Sq Mean Sq NumDF DenDF F.value Pr(>F)
## Type 551277 551277 1 3.024 3.9824 0.1392
## Canopy 67312 67312 1 39.899 0.4863 0.4896
## Type:Canopy 112517 112517 1 39.899 0.8128 0.3727
==> No significant effect. I am unsure what this tells us!
Part II. Leaflitter
1. Leaf fragmentation
==> Leaf fragmentation in agro streams one order of magnitude > than forest, no dif. between reaches (?check)
2- Decomposition
AFDM by stream type
## Canopy Type N Percent.AFDM.rem sd se ci
## 1 Closed Agro 28 23.31964 14.887963 2.813561 5.772949
## 2 Closed Forest 16 43.74562 9.175036 2.293759 4.889032
## 3 Open Agro 28 22.30214 13.913022 2.629314 5.394907
## 4 Open Forest 20 45.14400 11.604539 2.594854 5.431091
AFDM by mesh size
==> Decompositon in agro streams > than in forests, no differences between reaches (?check)
The preliminary model at the bottom suggests decomposition may be significnatly faster in nutrient reach streams. But to test this properly, we need to fit data to a non normal distribution (‘% AFDM remaining’ is proportion data, skewed -see humps and tails in quantile plot). Things I tried: log, arcsin transformations. We may need to restrict response to values 0-1.
A linear model using lmerTest, Type III Anova (unbal.design,normality assumption not met):
## Linear mixed model fit by REML t-tests use Satterthwaite approximations
## to degrees of freedom [lmerMod]
## Formula: Percent.AFDM.rem ~ Type * Canopy + (1 | Name)
## Data: Breakdown
##
## REML criterion at convergence: 711
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.7785 -0.5582 0.1432 0.7036 1.7044
##
## Random effects:
## Groups Name Variance Std.Dev.
## Name (Intercept) 42.25 6.50
## Residual 144.63 12.03
## Number of obs: 92, groups: Name, 5
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 33.6424 3.2374 3.0100 10.392 0.00188 **
## Type1 -10.9773 3.2374 3.0100 -3.391 0.04256 *
## Canopy1 0.2149 1.2963 85.3400 0.166 0.86873
## Type1:Canopy1 0.7392 1.2963 85.3400 0.570 0.57000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Type1 Canpy1
## Type1 -0.205
## Canopy1 0.030 -0.030
## Type1:Cnpy1 -0.030 0.030 -0.226## Analysis of Variance Table of type III with Satterthwaite
## approximation for degrees of freedom
## Sum Sq Mean Sq NumDF DenDF F.value Pr(>F)
## Type 1662.96 1662.96 1 3.009 11.4977 0.04256 *
## Canopy 3.97 3.97 1 85.338 0.0275 0.86873
## Type:Canopy 47.03 47.03 1 85.338 0.3252 0.57000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1==> So, there’s a sign. effect in decomposition that we need to understand better.
Our hypotheses were: 1) That biofilm accrual will be higher at open than at closed reaches==> Partially supported? No differences between reaches. We did find higher chlorophyll in nutrient-rich streams but this didn’t translate into more P uptake. 2) That higher biofilm accrual will boost litter breakdown at nutrient-poor but not at nutrient-rich streams ==> Not supported, as there wasn’t higher accrual in forests? or, what’s right way to test this? 3) That the effect will be higher for invertebrate fragmentation than for microbial decomposition ===> how do we see this? comparing fine vs coarse mesh?