Repository for Publications and Research Data
News from the ETH Library
Recently Added
Interaction of Environment and Vineyard Management Effects Shape Microbial Functionality, Terroir, and Grape Qualities
Item type: Journal Article
Flörl, Lena Victoria; Rueda-Mejia, Maria Paula; Reynard, Jean-Sébastien; et al. (2026)
Vineyards are complex agroecological systems, with geographically distinct microbial communities playing a central role in mediating vine health, grape quality, and ultimately local wine characteristics (terroir). However, the extent to which environmental factors and vineyard management practices interact to shape these microbial communities and their functions remains poorly understood. We implemented a multidisciplinary citizen science collaboration together with winegrowers to sample 19 Pinot Noir vineyards across two distinct Swiss regions (N = 38 samples). This allowed us to evaluate how a combination of environmental conditions and management practices (such as cover cropping and fungicide use) collectively impact soil and grape microbiomes, soil enzymatic activities, nutrient concentration in leaves, and grape chemistry. Regional differences were the primary drivers of variation, but multivariate analysis revealed a distinct, secondary influence from management practices on soil and berries. Berry quality was linked to functional soil properties, including microbial activity, which were further associated with key fermentative yeasts likely involved in wine fermentation. These findings highlight the interconnected nature of environmental, microbial, and agronomic factors in grapevine systems and offer novel insights into functional aspects of microbial terroir.
ESG Controversies in Global Firms: A Black Mark?
Item type: Journal Article
Bais, Beatrice; Orzes, Guido; Sartor, Marco (2026)
Despite increasing attention paid by companies to sustainability, there is still evidence of environmental, social and governance (commonly referred to as ESG) scandals. As research on this topic is scant, this paper aims to analyse the impact of ESG controversies on firms' sustainability practices, that is, ESG policies, as well as environmental management, communication and supply chain management actions. A heterogeneous sample of 1791 observations, comprising companies from various countries and sectors, is analysed using panel regressions over a 4-year period. Overall, our results seem to imply a rather weak or null effect of ESG controversy presence on firms' sustainability practices uptake decisions with no direct causality effect. In particular, firms do not appear to be deterred by negative ESG events when deciding whether to adopt ESG practices. Instead, we suggest that a range of external (e.g., market, stakeholder and political pressures) and internal factors (e.g., firm-level characteristics) jointly shape firms' ESG practice decisions. This work contributes to academic and practitioner knowledge as it resonates with companies' sustainability practices backsliding and approaches in order to make companies more responsible for their conduct.
Reducing biodiversity impacts of agricultural production while maintaining food supply – a modelling study of the Swiss food system
Item type: Journal Article
Augustiny, Eva; Müller, Adrian; Green, Ashley; et al. (2026)
Agri-food systems are essential for food security, yet they are major drivers of biodiversity loss through land conversion, intensification, and related emissions, threatening long-term stability of food production. To design food systems that sustain biodiversity while ensuring food provision, biodiversity impacts must be systematically assessed at food system level. This study integrates a food system model (SOLm) with biodiversity characterization factors from the Global Life Cycle Impact Assessment Method (GLAM), enabling consistent biodiversity impact assessment across multiple drivers. The analysis focuses on four drivers of biodiversity loss—land use, eutrophication, acidification, and climate change—and evaluates contributions to biodiversity impacts across alternative production scenarios in Switzerland. Results show that all alternative scenarios reduce biodiversity impacts from domestic production. Scenarios aligning livestock populations with grassland availability and reducing feed-food competition achieved the most substantial reductions, driven by lower pressures from climate change, eutrophication, and acidification, while maintaining calorie and protein supply. Land-use impacts decreased less, particularly in organic systems, due to lower yields, risking externalization of biodiversity impacts through increased imports. Impacts from imports were dominated by products such as coffee and cocoa, which exert substantial external pressures, although biodiversity impacts per unit area were lower for organic production than conventional. Thus, lower yields in organic systems necessitate additional measures, such as reducing feed-food competition and food loss and waste, to improve biodiversity outcomes. Coupling production-side measures with food supply shifts and food loss and waste reductions can lower biodiversity impacts without compromising food availability.
The proposed model for integrating biodiversity indicators with food system modelling provides a globally applicable framework for analysing trade-offs between food supply and biodiversity. However, due to lack of data, global biodiversity indicators lack relevant aspects, e.g., related to soil microbiome or eco-toxicity from plant protection means. Assessing significance and potential bias from missing aspects is required to ensure balanced conclusions.
Quantizing Recursive Reasoning Models
Item type: Working Paper
Ingolfsson, Thorir Mar; Tahir, Wajeeha; Tegon, Anna; et al. (2026)
Recursive reasoning models solve hard puzzles by applying compact, weight-tied blocks over many refinement steps. Because these blocks are reused many times, quantizing them creates a unique dynamical problem: the quantization error is incurred at every step. While 8-bit quantization (integer or float) preserves accuracy, moving to a per-tensor 4-bit format causes a systematic bias to accumulate. The ensuing drift catastrophically degrades exact-solution accuracy on Sudoku from 84.% to 0.0% (only 25% of cells correct). In this work, we show that this collapse is caused by activation-scaling \emph{granularity} rather than bit-width or number format. Crucially, moving to per-block scaling completely restores the transition. To implement this, we apply MXInt4, a blockwise integer activation format, to recursive reasoning models. It is competitive with blockwise float formats on our tasks, while keeping integer elements and power-of-two block scales. Finally, recursion depth and reuse modulate quantization sensitivity, with the deepest architecture we test (the EqR equilibrium model) the most sensitive. Yet blockwise scaling overcomes this vulnerability, staying robust across these architectures and transferring to the open-ended ARC-AGI benchmark.
The isotopic composition of solid-phase extracted dissolved organic nitrogen in seawater
Item type: Journal Article
Ryu, Yeongjun; Marconi, Dario; Smart, Sandi M.; et al. (2026)
Dissolved organic nitrogen (DON) is likely comprised of multiple compound classes with varying reactivities and turnover times, resulting in numerous roles in ocean biogeochemistry. Here, we present measurements of total DON and solid-phase extracted DON (SPE-DON) concentrations and δ¹⁵N values from sampling sites across the global ocean. An optimized SPE protocol was developed to maximize total DON recovery, with a recovery of 41.0 +/- 9.4% for surface DON across the global ocean and 56.6 +/- 11.2% for deep water DON from the Sargasso Sea. SPE-DON concentrations were 1.8-1.9 mu M across most sampling sites, in contrast to greater variation in total DON concentration (4.0-5.9 mu M). However, in the equatorial upwelling zones, SPE-DON concentrations were slightly (0.3-0.4 mu M) higher than in other regions. The δ¹⁵N of total DON in surface waters correlated well with the δ¹⁵N of nitrate supplied to the euphotic zone from the subsurface. SPE-DON δ¹⁵N was also correlated with nitrate δ¹⁵N, but SPE-DON δ¹⁵N values were confined to a narrower range compared to those of total DON. The combined concentration and δ¹⁵N data indicate that while SPE-DON is biased toward long-lived DON, it still retains some reactive components introduced through regional inputs in the upper ocean, and even some of its longer-lived components may be labile on the timescales of deep ocean circulation. The longer average turnover time of SPE-DON suggests that greater molecular polarity and/or charge directly increase or are otherwise correlated with the biogeochemical lability of different DON pools.