Impact of Dryocosmus kuriphilus on Castanea sativa Leaf Physiology and Morphology

Author: Konstantina Gianniou
Institution: Department of Environment, University of the Aegean
Study site: University Forest of Taxiarchis, Halkidiki, Greece (40°44'35"N, 23°18'12"E, 760 m a.s.l.)
Year: 2021

⚠️ Data & Code Disclaimer

\*\*The data in this repository are entirely synthetic.\*\*

The original field measurements were collected as part of an internship within the University of the Aegean and remain the intellectual property of the University of the Aegean, ELGO Demeter (Forest Research Institute) and Aristotle University of Thessaloniki. They are not included here.

The synthetic dataset (Analisi\\\_Sfika\\\_synthetic.csv, Analisi\\\_Physiologia\\\_synthetic.csv) was generated using generate\\\_synthetic\\\_data.R, which reproduces the \*\*statistical structure\*\* of the original data — including group sample sizes, provenance-level means and standard deviations, NA patterns, and inter-variable correlation structure — without exposing any real measurements.

\*\*The analysis code has also been rewritten from scratch\*\* to avoid reproducing the original internship report verbatim. The statistical methods, test choices, and interpretation are faithful to the original study; the implementation is new.

---

Overview The Asian Chestnut Gall Wasp (Dryocosmus kuriphilus Yasumatsu, 1951; Hymenoptera: Cynipidae) is an invasive species endemic to China that reached Europe via Italy in 2002, and was first recorded in Greece in 2014. It induces characteristic galls on chestnut (Castanea sativa Mill.) buds and shoots, causing leaf deformation, reduced photosynthetic capacity, and declines in flowering and fruiting. This study quantifies the impact of D. kuriphilus on the physiology and morphology of C. sativa leaves across five provenances from three countries (Greece, Italy, Spain) grown together in a common-garden plantation, eliminating confounding environmental variation.

Research Questions

1. Does infestation severity (DCI index) differ significantly between provenances or countries of origin?
2. Do reactivated dormant buds (RDB), gall counts, and dead shoots differ between provenances?
3. Are physiological leaf traits (CCI, A_sat, Fv/Fm, PI) affected by provenance or infestation level?
4. Which morphological traits (LMA, LDMC, leaf area, weight) are associated with infestation severity?
5. What is the multivariate structure of relationships between leaf traits and DCI?

---

Study Design Sampling

44 trees randomly selected from a common-garden plantation of 143 individuals Provenances: Coruna (Spain, n=12), Hortiatis (Greece, n=9), Malaga (Spain, n=6), Pellice (Italy, n=6), Sicily (Italy, n=11) For each tree: one branch (≥50 cm) + two leaves collected (July–August) Infestation index The DCI (Damage by Cynipid Infestation) index was computed following Gehring et al. (2018):

DCI = (Sd × 0.479 + Bdor × 0.525 + Gons × 0.120) × 100

where:

Sd = dead shoots / total shoots Bdor = reactivated dormant buds / total green shoots Gons = galls on sprouts / total green shoots

Leaf measurements

CCI --> Chlorophyll Content Index (field measurement)

A_sat --> Saturated photosynthesis rate (μmol m⁻² s⁻¹)

Fv/Fm --> Maximum quantum yield of PSII (chlorophyll fluorescence)

PI --> Performance Index

Gs --> Stomatal conductance

F_W / T_W / D_W --> Fresh, saturated, and dry leaf weight (g)

L_Area --> Leaf area (mm²) — scanned and digitised

LMA --> Leaf Mass per Area = D_W / L_Area (g mm⁻²)

LDMC --> Leaf Dry Matter Content = D_W / T_W

---

Statistical Methods

Infestation indices (DCI, RDB, Galls, Dead Shoots)

One-way ANOVA to test for differences between provenances and countries

Levene's test for homogeneity of variances (assumption check)

Shapiro-Wilk test on ANOVA residuals (normality check)

Kruskal-Wallis test as non-parametric backup when residuals are non-normal

Leaf physiological traits (CCI, A_sat, Fv/Fm, etc.)

Same ANOVA + Levene + Shapiro-Wilk + Kruskal-Wallis pipeline

Welch's ANOVA (oneway.test) when Levene's test indicates unequal variances

Tukey HSD post-hoc for significant ANOVAs

Correlations

Shapiro-Wilk to select correlation method

Spearman's rho (primary): non-parametric, robust to non-normality and outliers

Kendall's tau: reported alongside Spearman for key pairs with small n

Multivariate analysis

Principal Component Analysis (PCA) via FactoMineR::PCA() on tree-level averages

Two PCAs: (1) full trait set including gas exchange; (2) reduced set (PI, CCI, A_sat, LMA, LDMC, DCI)

Biplots produced with factoextra::fviz\\\_pca\\\_biplot(), coloured by provenance

Mixed-effects models

lmer(trait \\\~ Variety + DCI + LMA + (1|ID)) — tree ID as random intercept to account for the two-leaves-per-tree structure (pseudoreplication)

Fit via lme4; p-values from lmerTest (Satterthwaite approximation)

---

Key Findings (from the original study)

DCI by provenance --> No significant differences (ANOVA p = 0.6; KW p > 0.05)

CCI by provenance --> Significant (ANOVA p = 0.014); Hortiatis < Coruna (Tukey HSD)

CCI by country --> Significant (ANOVA p = 0.01); Spain > Greece

A_sat by provenance --> Significant (KW p = 0.008); Coruna > Hortiatis & Malaga

LDMC by provenance --> Marginal (KW p = 0.03); Coruna < Hortiatis

DCI ~ F_W --> Significant positive correlation (Spearman rho = 0.23, p = 0.04)

DCI ~ D_W --> Significant positive correlation (rho = 0.23, p = 0.03)

DCI ~ LMA --> Significant positive correlation (rho = 0.30, p = 0.005)

RDB ~ DCI --> Strong positive correlation (rho = 0.75, p < 0.001)

---

Repository Structure

dryocosmus-castanea/
├── generate\\\_synthetic\\\_data.R     ← Run this first to create the datasets
├── analysis.R                    ← Full statistical pipeline
├── fig1\\\_DCI\\\_provenance.png       ← DCI violin-boxplot by provenance
├── fig2\\\_RDB\\\_provenance.png       ← RDB boxplot by provenance
├── fig3\\\_CCI.png                  ← CCI by provenance and country
├── fig4\\\_PCA\\\_full.png             ← PCA biplot (all traits)
├── fig5\\\_PCA\\\_reduced.png          ← PCA biplot (reduced trait set)
├── fig6\\\_correlation\\\_matrix.png   ← Spearman correlation matrix
└── README.md

---

How to Run

# Step 1: generate the synthetic datasets
source("generate\\\_synthetic\\\_data.R")

# Step 2: run the full analysis (installs missing packages automatically)
source("analysis.R")

Requires R ≥ 4.1. All packages install automatically on first run.

References Gehring E., Bellosi B., Quacchia A. & Conedera M. (2018). Evaluating Dryocosmus kuriphilus-induced damage on Castanea sativa. Journal of Pest Science, Springer. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Contact Konstantina Gianniou
g.tem2106@gmail.com | LinkedIn |