The greenhouse experiment was carried out in the ScreenHouse phenotyping station, in the PhyTec Experimental Greenhouse, at the Institute of Bio- and Geosciences, Plant Sciences (IBG-2), Forschungszentrum Jülich GmbH, Germany (50°54′36″N, 6°24′49″E). This phenotyping platform has been already well-described by Nakhforoosh et al. (2016) and Scharr et al. (2017).

The experiment included five durum wheat genotypes (Duilio, Grecale, Iride, Marco Aurelio, and Saragolla). The following four treatments were performed: B1, with non-activated biochar from wood chips; B1D, with digestate-activated biochar from wood chips; B2, with non-activated biochar from wheat straw; B2D, with digestate-activated biochar from wheat straw. These treatments have been tested in relation to a negative control (C-), corresponding to the soil substrate (SS) lacking any biochar treatment. The control pots (C-) were filled only with 90% SS and 10% silica sand (expressed as dry weight percentages), previously mixed thoroughly in a mechanical mixer. The sample pots for the different biochar treatments were filled with 80% SS, 10% biochar (either pure or previously incubated with digestate), and 10% silica sand. Silica sand has been added to increase drainage. Each sample was replicated 6 times, resulting in 150 potted plants in total, using one plant per pot (Figures 1A,B). All the pots (5 L, 23 cm top diameter, 17 cm base diameter, 18 cm depth) were arranged in a completely randomized factorial design in the ScreenHouse.

Microclimate inside the greenhouse was monitored by sensors for relative humidity (RH, in%), temperature (T, in °C), and photosynthetic active radiation (PAR, expressed as photosynthetic photon flux density, in μmol⋅m/s). Supplemental light was provided to ensure 400 microE/m² at plant level during the day when incoming natural radiation was not sufficient.

Five Italian durum wheat varieties, commonly cultivated in the Peninsula for their high yield and remarkable commercial impact, were used in this study: Duilio and Marco Aurelio were kindly provided by Società Italiana Sementi (SIS), and Grecale, Iride, and Saragolla were kindly provided by Società Produttori Sementi, which is now Syngenta. Table 1 reports the information provided by the respective seed companies on major qualitative and morpho-physiological traits of interest of these five durum wheat varieties.

The soil substrate (SS) used in all the samples was a commercially available mixture of peat, sand, and pumice (namely SoMi 513, Dachstaudensubstrat; Hawita, Vechta, Germany; Table 2, left).

Two types of biochar obtained from two very contrasting feedstocks were used for the treatments in this study: B1 from wood chips and B2 from common wheat straw (Figures 1C,D, respectively); both were pyrolyzed in a “Schottdorf”-type reactor, and were kindly provided by Carbon Terra GmbH (Wallerstein, Germany). In particular, B1 was taken from the company’s GMP standard production, providing a certified biochar for animal feed. During its pyrolysis, biomass is first dried and then heated continuously up to 800°C for 36 h in an oxygen-free atmosphere; then, it reaches the oxidation zone, where 15% of the material is burned off and the rest falls through the grid. Thereafter, the resulting biochar is sprayed with water to stop the process, and leaves the system with approximately 20% humidity. Regarding B2, the straw is put into a 2-m³ steel box, tightly covered, and heated up to 750°C for 8 h; then it is sprayed with water to stop the process.

B1 is a reproducible type of biochar with homogeneous quality, and its composition and production procedure are well described. Analytical parameters of B1 are reported in Supplementary Table 1A of Kammann et al. (2015).

Here, the main chemical-physical properties of both B1 and B2 were again assessed (Table 2, left). Both types of biochar had alkaline pH (B1 8.5, B2 9.7). According to its woody feedstock, B1 showed > 4 times higher bulk density values, meaning lower porosity, than B2. Their electrical conductivity (EC) in water differed greatly, with that of B2 being > 8 times higher than that of B1 (Table 2, left).

In order to investigate their effects on plants, the two types of biochar were added to the SS either in their pure (B1 and B2 treatments) or previously activated form, i.e., spiking with nutrients, using digestate (B1D and B2D treatments). In practice, for nutrient spiking, a digestate from a commercial biogas facility operating with maize silage was used, as described in previous studies (Robles-Aguilar et al., 2019; Dietrich et al., 2020). Briefly, the fresh digestate used consisted of 7.2% dry matter and 5.3% organic substance. It contained 0.53% N (of which 0.32% was ammonium-N), with a C/N ratio of 6. Furthermore, 0.14% phosphorus, 0.68% potassium, 0.037% magnesium, 0.16% calcium, and 0.03% sulfur were detected. For the activation process, each biochar was wrapped into a stable permeable tissue and submerged in a 60-L bin filled with digestate, allowing the liquid, nutrient-rich fraction to penetrate and soluble nutrients to be absorbed by the biochar (Dietrich et al., 2020; Figure 1E). After 10 days of incubation, each biochar was partially open air dried overnight (Figure 1F). After the treatment with digestate, both B1D and B2D showed lower dry substance content (−37.5 and −63.3% for B1 and B2, respectively), and low pH reduction (<8.3%) than the pure biochar (Table 2, left). The digestate treatment had a minor opposite effect on dry bulk density, which showed slighter reduction (−14.8%) in B1D than in B1, and increase (+34.1%) in B2D compared with B2 (Table 2, left). N content (mainly present in the ammonium form) increased considerably in both nutrient-spiked types of biochar, from less than 2 mg/L in both types of biochar up to 31 mg/L in B1D and 308 mg/L in B2D. In a similar way, the content of main macronutrients also increased notably. In particular, P was > 79.8%, and K was > 31.5% in both nutrient-spiked types of biochar. Concerning Mg, its increase was very high (81%) in B2D and more moderate (13.4%) in B1D (Table 2, left).

To supply the pots with precise amounts of each component (SS, silica sand, B1, B1D, B2, and B2D), their moisture content was measured with HB43-S Moisture Analyzer (Mettler Toledo, Gießen, Germany). Soil analyses of the potting components before mixing with each other (Table 2, left) and of the potting mixtures of the control and each treatment (B1, B1D, B2, and B2D) at the beginning of the experiment (before seed germination, T₀) as well as the end (T_f) after 58 days (Table 2, center and on the right, respectively), were performed according to the procedures applied by LUFA NRW, Landwirtschaftskammer Nordrhein-Westfalen (VDLUFA, 1991).

Three seeds of uniform size and weight per pot were put to germinate directly in the soil substrate, suitably spaced from each other 3 cm below the air-soil interface. After 3 days, only one plantlet per pot was kept, and the other two were removed. The pots were irrigated three times per week to keep soil moisture level around a 50% water holding capacity throughout the experiment with an automated watering system. Pot water loss due to plant evapotranspiration (ET) was recorded by weighing the pots before watering, ensuring equal soil moisture levels. Given the relatively high percentage of humidity inside the ScreenHouse, we assumed that most of the weighed water loss was related to plant transpiration.

The imaging station in the ScreenHouse allowed capturing data on leaf area expansion, inferred from the number of green pixels in the image belonging to the plant (Golzarian et al., 2011). The phenotyping system allowed repeated measures over the projected shoot system area (PSSA). In correspondence with the ET measures, each plant was imaged (RGB) for dynamic estimation of shoot biomass (as projected shoot system area, PSSA) three times per week. After each measurement, the pots were automatically re-randomized with a laser positioning system and a robotic crane to avoid any systematic bias from position within the greenhouse. The subsequent imaging processing pipeline has been described earlier by Nakhforoosh et al. (2016). Plant growth was evaluated in the early vegetative stage during the course of the experiment for a total of 58 days until the experiment was stopped (Figure 1G).

Additional plant traits were measured to aid plant behavior evaluation (Table 3). Plant phenological developmental stage was observed throughout the experiment and evaluated by BBCH-scale (Meier, 1997), acronym for Biologische Bundesanstalt für Land- und Forstwirtschaft, Bundessortenamt und CHemische Industrie, 7 weeks after seed germination and at the end of the experiment. At the end of the experiment (T_f), the number of tillers per plant (tiller number, TN), number of spikes per plant (spike number, SN), spike length (SL), and plant height (PH) from the base of the stem up to the end of the emerging spike were measured. After that, aboveground tissue was excised, and shoot fresh weight (FW) was annotated. Plant shoot area (PSA_Licor) was also measured with a LI-3100C Area Meter (LI-COR, Inc., Bad Homburg, Germany), and then shoot dry weight (DW) was finally recorded after 48 h of drying at 75°C. Actual water content (WC) of the aboveground plant body was calculated by the formula WC = 100*(FW-DW)/FW.

All the statistical analyses were conducted with the IBM SPSS Statistics 23 software. Data were analyzed for their normality and equality of variances by the Shapiro-Wilk test and the Levene test, respectively. When these two conditions were assessed, a two-way ANOVA was carried out for the dependent variables, represented by the plant-related traits, with “genotype” and “treatment” as fixed factors (independent variables). The Ryan-Einot-Gabriel-Welsch-and-Quiot (REGWQ) method with Bonferroni correction was used for post hoc testing. Differently, when a dataset did not meet the criteria of the equality of variances, a one-way ANOVA was performed with Welch correction and the Games-Howell method for post hoc testing. The selected statistical significance, depending on the data and test, is reported case by case.

A bivariate Pearson correlation analysis was conducted for the plant trait datasets, after assessing that they did not violate the assumptions of data normality and homoscedasticity, with a two-tailed test and p < 0.01. In a different way, the correlation between the two time-series datasets of plant ET and PSSA was examined with the non-parametric Spearman coefficient.

When observing the potting soil mixtures tested (SS, i.e., C-, B1, B1D, B2, and B2D), and comparing their main parameters at the beginning (T₀) and the end of the experiment (T_f) (Table 2, middle and on the right, respectively), C- showed the highest dry bulk density values among all the biochar treatments, which means it has low porosity. As a general behavior, in the analyzed samples, while dry bulk density remained almost constant, the dry substance tended to decrease during the experiment. A relevant exception to this trend was represented by the B2 mixture, which showed the lowest values of dry substance content (47.4% at T₀ that differently from the general trend increased up to 58.1% at T_f) and dry bulk density (371 g/L at T₀ that did not vary considerably at T_f, as for the other treatments).

As usual, the addition of biochar in the soil substrate resulted in pH increase more pronounced in the mixture with pure biochars than in those with digestate-treated biochars. There were no marked pH changes from the beginning to the end of the experiment, and all the SSs and their mixtures used could be classified as neutral or weak alkaline, with values ranging from 6 to 6.5 in C- and up to 7.5 to 7.6 in B2 (Table 2). Electrical conductivity (EC) in water, and KCl salt in water as well as in calcium sulfate showed a general decrease at the end of the experiment, but for these traits there was again the exception of the B2 mixture, whose starting values were always lower than in the other soil mixtures, and different from the general fashion they increased considerably at the end (in particular, EC in water of B2 was 40.8% lower than C- at T₀ and 51.6% higher than C- at T_f, increasing 1.7 times during the experiment duration). It is also noticeable that the highest values of these three traits were detected in the mixture B2D, both at T₀ and at T_f (in particular, EC in water of B2D was 42.8% higher than C- at T₀ and 35.2% higher than C- at T_f, decreasing 1.4 times during the experiment duration), respectively, 841 and 596 μS/cm for EC, 2.95 and 2.31 g/L for KCl salt in H₂O, and 2.34 and 1.71 for KCl salt in CaSO₄; Table 2).

The main nutrients (N, P, K, and Mg) showed obvious reduction at the end of the experiment with respect to T₀, with few exceptions. Unexpectedly, in B2, an increase in nitrate-nitrogen and in potassium (K) was detected from T₀ to T_f. It is also noticeable that the soil mixtures with wheat straw biochar, either as pure (B2) or incubated in the digestate (B2D), resulted in higher concentration of phosphorous (from 2 up to 3.2 times more than in C-) and K (from 3.7 up to 6.7 times more than in C-) (Table 2).

Plant evapotranspiration (ET) was estimated as ml of water loss by weighting the pots throughout the course of the experiment; As observed for the PSSA plant trait, ET also showed an overall increasing dynamic trend, but with relevant fluctuations shaping a zigzag distribution (Figures 2C,D), possibly related to the inter-day smooth variations under the microclimate conditions (T, RH, and PAR) inside the ScreenHouse (Supplementary Material 3). At the beginning of the experiment, the daily ET from each pot was around 50 ml in all the samples, and then it increased, as expected, with the growth of the plants and extension of plant surfaces available for the transpiration process. The ET dataset on day 56, close to the end of the experiment, was analyzed by two-way ANOVA, evidencing statistically significant differences due to “genotype” as well “treatment” (p <0.001), but not due to the “genotype*treatment” interaction (Table 5 and Supplementary Material 4). Focusing on “genotype,” ET_d56 in Duilio was significantly higher (p <0.001) than in Grecale and Marco Aurelio (>22.5 and >17.7%, respectively) and in Iride and Saragolla, which even had lower significance (p <0.05). Focusing on “treatment,” B2 negatively affected ET (36% lower than the C-), and indeed it was significantly different from the control and all the other samples (p <0.001); ET in B2D was significantly lower than that in B1 and B1D (p <0.001), and had lower probability than C- (p <0.05) (Figure 4).

The plant traits related to wheat morphology and yield-tiller number (TN), spike number (SN), spike length (SL), and plant height (PH) were measured at T_f to aid in plant behavior evaluation (Table 3). The results of the one-way ANOVA showed that no significant differences were observed in SN and in SL, and that TN showed differences among the genotypes and treatments. In particular, Marco Aurelio, among the genotypes, and B2, among the treatments, showed the lowest average tiller number (p >0.05, data not shown).

Concerning PH, resulting from the two-way ANOVA, a highly significant effect of “genotype” (p <0.001) and a slightly significant effect of “treatment” (p <0.05) was observed, while the interaction term “genotype*treatment” was not statistically significant (p >0.05) (Table 6). The post hoc results of PH data analysis are reported in the form of histogram in Figure 5A. With respect to the “genotype” factor, PH in Duilio was significantly higher than that in all the other genotypes, followed by Marco Aurelio; then, Grecale, Iride and Saragolla presented the lowest values, similar among each other. With respect to the “treatment” factor, B1 and B1D did not affect PH. Moreover, while B2 had a negative effect on PH, when incubated with digestate (i.e., B2D) the PH reduction was less pronounced (Figure 5A).

Additionally, the agro-physiological traits related to the aboveground plant tissues, fresh weight (FW), and dry weight (DW), besides water content (WC), were evaluated at T_f. The results of the two-way ANOVA are reported in Table 6. FW showed a highly significant effect of both “genotype” and “treatment” as well as the “genotype*treatment” interaction (p < 0.01); and the same was observed for DW, even though the interaction was more significant (p < 0.005). WC showed a significant effect of both “genotype” and “treatment,” but not of the interaction term. The resulting histograms from mean values and related errors of these traits are also represented in Figure 5. With respect to “genotype,” Duilio, Iride, and Saragolla showed a higher FW (35.3, 35.1, and 33.8 g, respectively) than Marco Aurelio and Grecale (27.2 and 26.7 g, respectively). With respect to “treatment,” B2 had a reliable 43.2% reduction in FW compared C-, which was slightly relieved (24.3%) by the digestate treatment in B2D, while the application of woody biochar (B1 and B1D) did not significantly affect FW with respect to C- (Figure 5B). In general, the DW results reflected those of FW, as expected, with Duilio (5.42 g) exhibiting the highest DW values, followed by Iride and Saragolla (4.92 and 4.85 g, respectively), and then by Grecale and Marco Aurelio (4.36 and 4.21 g, respectively) with the lowest DW values. At the same time, the DW results distributed similarly to FW also in relation to the treatments, with B2 being related to the lowest DW values (66.9% lower than C-), followed by B2D (14.5% lower than C-; Figure 5C).

Regarding WC, even though only low-entity significant differences were observed, Grecale and Marco Aurelio showed the lowest WC percentages together with Duilio (83.6, 83.7, and 84.3%, respectively), which for the first time appeared to cluster with these genotypes, while Saragolla and Iride presented a higher WC (85.2 and 85.8%, respectively). Concerning the treatments, B2 and B2D resulted in an inferior plant tissue hydration level (3.2 and 2.4% lower than C-, respectively; Figure 5D).

Bivariate Pearson correlation analysis was performed for the following plant variables at the end of the experiment: PSSA, PSA_Licor, ET, PH, FW, DW, and WC; the resulting correlation matrix is reported in Table 4. Besides PSSA with PSA_Licor, strong positive linear relationship correlations (p > 0.01) were ascertained among most of the analyzed plant phenotypic traits. The only exceptions were represented by FW with PH showing a positive correlation with lower significance (p > 0.05), and by WC with PH, whose result did not correlate at all. As it can be expected, ET resulted to be directly proportional to all the plant traits, except for PH that anyhow demonstrated no link with the other traits. Among the variables analyzed here, the lower moderate correlation was that between DW and WC.

A schematic representation summarizing the effects of both “genotype” and “treatment” on the analyzed plant phenotypic traits is shown in Table 7. The “genotype” effect is shown (Table 7, left) using Duilio as the reference, while the control was used as reference for summing up the “treatment” effect (Table 7, right). Accordingly, while Iride and Saragolla behaved like Duilio with respect to PSSA, PSA_Licor, ET, FW, and DW, always showing higher values for these traits, in Grecale and Marco Aurelio, these traits were negatively influenced. Duilio, in particular, often showed the highest values, for DW it even clustered apart with values significantly higher than those for Iride and Saragolla (p < 0.01). On the other hand, the wood biochar, both as pure and after nutrient spiking with digestate (B1 and B1D, respectively), behaved like the control in promoting plant growth, while the wheat straw biochar, both in B2- and in B2D-treated plants, negatively affected plant growth.

A positive correlation was also found in the measurement of ET and PSSA repeated over time (Figure 2), probably not related in a linear fashion, and resulting in a Spearman’s rho equal to 0.679 (p < 0.01, two-tailed; Supplementary Material 5).