Learning vs. understanding: When does artificial intelligence outperform process-based modeling in soil organic carbon prediction?

In recent years, machine learning (ML) algorithms have gained substantial recognition for ecological modeling across various temporal and spatial scales. However, little evaluation has been conducted for the prediction of soil organic carbon (SOC) on small data sets commonly inherent to long-term soil ecological research. In this context, the performance of ML algorithms for SOC prediction has never been tested against traditional process-based modeling approaches. Here, we compare ML algorithms, calibrated and uncalibrated process-based models as well as multiple ensembles on their performance in predicting SOC using data from five long-term experimental sites (comprising 256 independent data points) in Austria. Using all available data, the ML-based approaches using Random forest and Support vector machines with a polynomial kernel were superior to all process-based models. However, the ML algorithms performed similar or worse when the number of training samples was reduced or when a leave-one-site-out cross validation was applied. This emphasizes that the performance of ML algorithms is strongly dependent on the data-size related quality of learning information following the well-known curse of dimensionality phenomenon, while the accuracy of process-based models significantly relies on proper calibration and combination of different modeling approaches. Our study thus suggests a superiority of ML-based SOC prediction at scales where larger datasets are available, while process-based models are superior tools when targeting the exploration of underlying biophysical and biochemical mechanisms of SOC dynamics in soils. Therefore, we recommend applying ensembles of ML algorithms with process-based models to combine advantages inherent to both approaches.

Soil quality increases with long-term chabazite-zeolite tuff amendments in arable and perennial cropping systems.

The application of natural zeolites to improve soil quality and functioning has become highly popular, but we still miss information about the long-term effects on the soil due to its application. This study assesses the soil quality index (SQI) of three distinct agricultural soil systems 6-10 years after a single application of natural chabazite zeolite as a soil amendment. These soils exhibit different management practices: intensive arable (cereals), intensive perennial (pear) and organic perennial (olive). In the arable system, a zeolite application dosage of 5, 10 and 15 kg m-2 was tested and compared to unamended soil. In the two perennial systems, an application of 5 kg m-2 was tested against untreated reference sols. A set of 25 soil physical, chemical and biological parameters linked to soil health and quality were analysed at each experimental site. The dataset was investigated through Principal Component Analysis (PCA) to calculate the soil quality index (SQI) using linear scoring. In the arable-cereal field, the SQI doubled (0.3 to ca. 0.6 for all amendments) in chabazite-amended plots; a dose effect was not recognizable. In both perennial fields, the SQI was significantly higher in the chabazite-amended plots (5 kg m-2) with similar increases as compared to the arable-cereal field. At each site, the indicators selected by the PCA were different, indicating that chabazite addition impacted soil quality differently in each cropping system. Overall, the results highlighted a significant increase in soil quality with chabazite amendment, which confirms its potential for increasing soil health in the long-term.

Rewetting the hyper-arid Atacama Desert soil reactivates a carbon-starved microbial decomposer community and also triggers archaeal metabolism.

Extreme environmental conditions make soils of the hyper-arid Atacama Desert one of the most hostile habitats for life on the planet. During the short intervals of moisture availability that occur, it remains unresolved how soil microorganisms physiologically respond to such dramatic environmental changes. Therefore, we simulated a precipitation event - without (H2O) and with (H2O + C) labile carbon (C) supplementation - and investigated the responses in microbial communities (using phospholipid fatty acids (PLFAs) and archaeal glycerol dialkyl glycerol tetraether (GDGTs)) and physiology (by means of respiration, bacterial and fungal growth and C-use efficiency (CUE)) during a five-day incubation. We demonstrated that bacterial and fungal growth does occur in these extreme soils following rewetting, albeit at 100-10,000-fold lower rates compared to previously studied soil systems. C supplementation increased levels of bacterial growth and respiration responses by 5- and 50-fold, respectively, demonstrating a C-limited microbial decomposer community. While the microbial CUE following rewetting was c. 14 %, the addition of labile C during rewetting resulted in a substantial reduction (c. 1.6 %). Consistent with these interpretations, the PLFA composition clearly shifted from saturated towards more unsaturated and branched PLFAs, which could arise from (i) a physiological adaptation of the cell membrane to changing osmotic conditions or (ii) a community composition shift. Significant increases in total PLFA concentrations were solely found with H2O + C addition. Contrary to other recent studies, we found evidence for a metabolically active archaeal community in these hyper-arid soils upon rewetting. We conclude that (i) microorganisms in this extreme soil habitat can be activated and grow within days following rewetting, (ii) available C is the limiting factor for microbial growth and biomass gains, and (iii) that an optimization of tolerating the extreme conditions while maintaining a high CUE comes at the expense of very poor resource-use efficiency during high resource availability.

Stabilization of mineral-associated organic carbon in Pleistocene permafrost.

Ice-rich Pleistocene-age permafrost is particularly vulnerable to rapid thaw, which may quickly expose a large pool of sedimentary organic matter (OM) to microbial degradation and lead to emissions of climate-sensitive greenhouse gases. Protective physico-chemical mechanisms may, however, restrict microbial accessibility and reduce OM decomposition; mechanisms that may be influenced by changing environmental conditions during sediment deposition. Here we study different OM fractions in Siberian permafrost deposited during colder and warmer periods of the past 55,000 years. Among known stabilization mechanisms, the occlusion of OM in aggregates is of minor importance, while 33-74% of the organic carbon is associated with small, <6.3 µm mineral particles. Preservation of carbon in mineral-associated OM is enhanced by reactive iron minerals particularly during cold and dry climate, reflected by low microbial CO2 production in incubation experiments. Warmer and wetter conditions reduce OM stabilization, shown by more decomposed mineral-associated OM and up to 30% higher CO2 production. This shows that considering the stability and bioavailability of Pleistocene-age permafrost carbon is important for predicting future climate-carbon feedback.

Straw amendment and nitrification inhibitor controlling N losses and immobilization in a soil cooling-warming experiment.

It is common practice in agriculture to apply high­carbon amendments, e.g. straw, or nitrification inhibitors (NI) to reduce soil nitrogen (N) losses. However, little is known on the combined effects of straw and NI and how seasonal soil temperature variations further affect N immobilization. We conducted a 113-day mesocosm experiment with different levels of 15N-fertilizer application (N0: control; N1: 125 kg N ha-1; N2: 250 kg N ha-1) in an agricultural soil, amended with either wheat straw, NI or a combination of both in order to investigate N retention and loss from soil after a cooling-warming phase simulating a seasonal temperature shift, i.e., 30 days cooling phase at 7 °C and 10 days warming phase at 21 °C. Subsequently, soils were planted with barley as phytometers to study 15N-transfer to a following crop. Straw addition significantly reduced soil N-losses due to microbial N immobilization. Although carbon added as straw led to increased N2O emissions at high N fertilization, this was partly counterbalanced by NI. Soil cooling-warming strongly increased ammonification (+77 %), while nitrification was suppressed, and straw-induced microbial N immobilization dominated. N immobilized after straw addition was mineralized at the end of the experiment as indicated by structural equation models. Re-mineralization in N2 was sufficient, but still suboptimal in N0 and N1 at critical times of early barley growth. N-use efficiency of the 15N tracer decreased with fertilization intensity from 50 % in N1 to 35 % in N2, and straw amendment reduced NUE to 25 % at both fertilization rates. Straw amendment was most powerful in reducing N-losses (-41 %), in particular under variable soil temperature conditions, but NI enforced its effects by reducing N2O emission (-40 %) in N2 treatment. Sufficient N-fertilization coupled with straw application is required to adjust the timely re-mineralization of N for subsequent crops.

Recycling nitrogen from liquid digestate via novel reactive struvite and zeolite minerals to mitigate agricultural pollution.

Recycling nutrients is of paramount importance. For this reason, struvite and nitrogen enriched zeolite fertilizers produced from wastewater treatments are receiving growing attention in European markets. However, their effects on agricultural soils are far from certain, especially struvite, which only recently was implemented in EU Fertilizing Product Regulations. In this paper, we investigate the effects of these materials in acid sandy arable soil, particularly focusing on N dynamics, evaluating potential losses, transformation pathways, and the effects of struvite and zeolitic tuffs on main soil biogeochemical parameters, in comparison to traditional fertilization with digestate. Liming effect (pH alkalinization) was observed in all treatments with varying intensities, affecting most of the soil processes. The struvite was quickly solubilized due to soil acidity, and the release of nutrients stimulated nitrifying and denitrifying microorganisms. Zeolitic tuff amendments decreased the NOx gas emissions, which are precursors to the powerful climate altering N2O gas, and the N enriched chabazite tuff also recorded smaller NH3 emissions compared to the digestate. However, a high dosage of zeolites in soil increased NH3 emissions after fertilization, due to pronounced pH shifts. Contrasting effects were observed between the two zeolitic tuffs when applied as soil amendments; while the chabazite tuff had a strong positive effect - increasing up to ∼90% the soil microbial N immobilization - the employed clinoptilolite tuff had immediate negative effects on the microbial biomass, likely due to the large quantities of sulphur released. However, when applied at lower dosages, the N enriched clinoptilolite also contributed to the increase of microbial N. From these outcomes, we confirm the potential of struvite and zeolites to mitigate the outfluxes of nutrients from agricultural systems. To gain the best results and significantly lower environmental impacts, extension practitioners could give recommendations based on the soils that are planned for zeolite application.

Soil fertility relates to fungal-mediated decomposition and organic matter turnover in a temperate mountain forest.

Fungi are known to exert a significant influence over soil organic matter (SOM) turnover, however understanding of the effects of fungal community structure on SOM dynamics and its consequences for ecosystem fertility is fragmentary. Here we studied soil fungal guilds and SOM decomposition processes along a fertility gradient in a temperate mountain beech forest. High-throughput sequencing was used to investigate fungal communities. Carbon and nitrogen stocks, enzymatic activity and microbial respiration were measured. While ectomycorrhizal fungal abundance was not related to fertility, saprotrophic ascomycetes showed higher relative abundances under more fertile conditions. The activity of oxidising enzymes and respiration rates in mineral soil were related positively to fertility and saprotrophic fungi. In addition, organic layer carbon and nitrogen stocks were lower on the more fertile plots, although tree biomass and litter input were higher. Together, the results indicated a faster SOM turnover at the fertile end of the gradient. We suggest that there is a positive feedback mechanism between SOM turnover and fertility that is mediated by soil fungi to a significant extent. By underlining the importance of fungi for soil fertility and plant growth, these findings furthermore emphasise the dependency of carbon cycling on fungal communities below ground.

Non-structural carbohydrate concentrations of Fagus sylvatica and Pinus sylvestris fine roots are linked to ectomycorrhizal enzymatic activity during spring reactivation.

We evaluated whether changes in fine root non-structural carbohydrate reserves of Fagus sylvatica and Pinus sylvestris trees influence potential enzymatic activities of their ectomycorrhizal symbionts from winter towards spring reactivation, and whether these changes influence potential soil enzymatic activities. We analyzed sugar and starch concentrations in the fine roots of Fagus sylvatica and Pinus sylvestris and potential activities of ß-glucosidase, ß-xylosidase, and cellobiohydrolase (as proxies for carbon-degrading enzymes) as well as leucine aminopeptidase and chitinase (as proxies for nitrogen-degrading enzymes) of their dominant ectomycorrhizal symbionts as well as in the soil. Sugar concentrations in the fine roots were significantly positively correlated with enzymatic activities of the ectomycorrhizal symbionts. In Pinus sylvestris, both carbon- and nitrogen-degrading enzyme activities showed significant positive correlations with fine root sugar concentrations. In Fagus sylvatica, fine root sugar concentrations were explicitly positively correlated with the activity of nitrogen-degrading enzymes. The chitinase activity in the soil was found to be strongly positively correlated with the enzymatic activity of the ectomycorrhizal symbionts as well as with fine root sugar concentrations. Fine root carbohydrate concentrations of Fagus sylvatica and Pinus sylvestris trees and enzymatic activities of their associated ectomycorrhizal fungi are connected. The specific nutrient demand of the tree species during spring reactivation may affect ectomycorrhizal enzymatic activity via carbon mobilization in the fine roots of Fagus sylvatica and Pinus sylvestris. Moreover, our results suggest that trees indirectly contribute to the degradation of fungal necromass by stimulating ectomycorrhizal chitinase activity in the soil.