This dataset contains the raw files and R code describing the pre-processing and analysis of a common garden experiment used to quantify the plasticity and adaptive responses to seasonal and geographic climate variation of Nannochloropsis, a microalga commonly used in biotechnology. An initially monoclonal strain was grown outdoors across four locations in Hawaii (Cyanotech), Texas (Qualitas Health), California (UCSD), and New Mexico (NMSU). Following 17 and 22 months of cultivation outdoors, we collected samples during winter and summer, respectively, and we compared strains’ growth from the four sites across temperature and light gradients in the laboratory.

Experimental design: We tested the performance of each strain compared to the baseline strain in 24-well deep well plates (VWR, Radnor, PA, USA) exposed to temperature (5°C, 10°C, 15°C, 20°C, 25°C, 30°C) and light (100% = 100 µmol m^-2 s^-1, 50% = 50 µmol m^-2 s^-1, 10% = 10 µmol m^-2 s^-1 and 5% = 5 µmol m^-2 s^-1) gradients.

Each strain was inoculated in three replicate wells 24 times (for each treatment combination of light and temperature). We randomized the position of the plates and replicates in plates and used sterile distilled water in well A1 of each plate, as well as culture media in empty wells to confirm there was no cross-contamination across wells. We collected samples for optical density on day 0, 1, 3 and 5. OD₇₅₀ was measured on 1:1 diluted sample using a Tecan Infinite 200 PRO microplate reader (Tecan, Männedorf, Switzerland).

Calculation of growth rates and statistical analyses: Cultures were grown for 5 days, and their growth rates calculated from the slope of OD₇₅₀ increase during the exponential growth phase (i.e., the first 3 days of growth) and used as our measure of performance. We constructed thermal and light performance curves (TPC and LPC) respectively using generalized additive models (GAMs) from the package “mgcv”. We quantified the effects of the site (baseline, Hawaii, Texas, California, and New Mexico) and seasonality (winter and summer) on growth rate by comparing models with site-specific or global TPCs/LPCs. For TPCs, each light treatment was analyzed separately, whereas for LPCs, each temperature treatment was analyzed separately. To select for the best fitting model, we used Restricted Maximum Likelihood (REML). Within each light/temperature treatment and using season or site as the grouping parameters, we applied GAMs comparing three nested models, (i) Model 1: the “simplest” model using a single global smoother fitting all data; (ii) Model 2: a model with a global smoother and a parametric term (i.e., site or season), allowing different intercepts for each group (site or season); and (iii) Model 3: a model providing different smoothers for each group and a parametric term for each group-specific intercepts. Models were selected based on the lowest Akaike information criterion (AIC). The anova.gam function performed Wald tests to determine the significance of each parametric and smooth terms. We avoided overfitting data and made sure that every model had a gaussian (bell shape) curve as expected for TPC/LPC when possible, by limiting wiggliness. We used R version 4.2.2 for statistical analyses. All differences were considered significant when p < 0.05.