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

New Biotechnology, Journal Year: 2024, Volume and Issue: 81, P. 20 - 31

Published: March 8, 2024

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

Language: Английский

---

VIS-NIR spectroscopy and environmental factors coupled with PLSR models to predict soil organic carbon and nitrogen

International Soil and Water Conservation Research, Journal Year: 2024, Volume and Issue: 12(4), P. 844 - 854

Published: Feb. 14, 2024

Soil profile organic carbon (OC) and total nitrogen (TN) are influenced by topographic attributes, land use. The visible near-infrared (Vis-NIR) spectroscopy method can be used for the prediction of OC TN because it is reliable, nondestructive, fast, cost-effective. VIS-NIR soil spectral environmental data were combined with Partial least squares regression (PLSR) model to examine effect topography attributes use on topsoil subsoil stocks. After this, based depth, 114 samples collected from 0 20 cm (topsoil) 20–50 (subsoil) under three uses, as well TN, along several properties including particles (sand, silt, clay), pH, bulk density in both measured. A DEM a resolution 30 m was derive factors remote sensing calculate vegetation index. Soils (0–50 cm) orchard had highest stock SOC (7.4 kg m−2) (2.4 m−2). There significant increase matter soils located south aspect (8.3 compared other aspects, particularly north (3.9% increase). contain higher soil-water contents lower temperatures, resulting decrease decomposition matter. strong positive correlation demonstrated between wetness index (0.57–0.63) stocks (0.54–0.66) loading score among terrain suggesting that primary factor controlling stocks, Additionally, we found south-facing aspects (N aspects) spectra. PLSR, which showed an R2 0.82, RMSE 0.15 %, RPD 0.39 indicated excellent capabilities content. We concluded PLSR coupled Vis-NIR able predict N content different slopes.

Language: Английский

---

Benchmarking carbon sequestration potentials in arable soils by on-farm research on innovative pioneer farms

Plant and Soil, Journal Year: 2022, Volume and Issue: 488(1-2), P. 137 - 156

Published: Aug. 1, 2022

Abstract Purpose Tackling the global carbon deficit through soil organic (SOC) sequestration in agricultural systems has been a focal point recent years. However, we still lack comprehensive understanding of actual on-farm SOC potentials order to derive effective strategies. Methods Therefore, chose 21 study sites North-Eastern Austria covering wide range relevant arable types and determined pool sizes (0–35 cm depth) pioneer versus conventional management relation permanently covered reference soils. We evaluated physico-chemical predictors stocks quality differences between using Fourier-transform infrared (FTIR) spectroscopy. Results Compared farming systems, were 14.3 Mg ha − 1 or 15.7% higher equaling rate 0.56 yr . Reference soils however showed approximately 30 50% than respectively. Nitrogen dissolved similar patterns. While could close storage coarse-textured soils, medium- fine-textured 30–40% lower compared quality, as inferred by FTIR spectra, differed land-use yet lesser extent cropping systems. Conclusions Innovative alleviates storage. Actual realized are rather estimated derived from field experiments models. The potential is governed parameters. More approaches necessary evaluate close-to-reality agroecosystems.

Language: Английский

---

Predictors for digital mapping of forest soil organic carbon stocks in different types of landscape

Soil and Water Research, Journal Year: 2022, Volume and Issue: 17(2), P. 69 - 79

Published: Feb. 9, 2022

Forest soils have a high potential to store carbon and thus mitigate climate change. The information on spatial distribution of soil organic (SOC) stocks is very important. This study aims analyse the importance environmental predictors for forest SOC stock prediction at regional national scale in Czech Republic. A big database data more than 7 000 sites was compiled from several surveys. were calculated content bulk density topsoil mineral layer 0-30 cm. Spatial models developed separately individual natural areas four subsets with different altitude range, using random method. strongly differs between regions altitudes. At lower altitudes, edaphic series classes are strong predictors, while higher altitudes related topography become depends pedodiversity level difference classes. contribution types as limited when one (mostly coniferous) type dominates. Better results can be obtained smaller, but consistent regions, like some areas.

Language: Английский

---

Letter to the Editor regarding Feeney et al., 2024, ‘Benchmarking soil organic carbon (SOC) concentration provides more robust soil health assessment than the SOC/clay ratio at European scale’

The Science of The Total Environment, Journal Year: 2025, Volume and Issue: 976, P. 179278 - 179278

Published: April 7, 2025

Language: Английский

---

Evaluating Natural Climate Solutions in Long-Term Climate Strategies: Opportunities for Enhanced Mitigation Across the European Union

Land, Journal Year: 2025, Volume and Issue: 14(4), P. 825 - 825

Published: April 10, 2025

Effective national strategies must be carefully planned in advance to position the land sector as a pivotal contributor achieving 2050 climate neutrality target set by European Union (EU) under Paris Agreement. Governments define their pathways achieve goals through long-term low emission development (LTSs), which describe policies and measures for just socially fair transition greenhouse gas. This paper explores natural solutions foreseen available twenty-five LTSs EU assess each country’s use potential of mitigation. Subsequently, it evaluates if better planning is possible increase carbon storage reduce gas emissions. We found that most analyzed countries envisage wide solutions, demonstrating an understand importance meet targets. Sustainable forest management, agroforestry, healthy sustainable diets emerge mostly reported with largest mitigation across EU, albeit large variability among countries. However, some could further harness sector. study highlights how improve LTSs, accounting specific potential.

Language: Английский

---

Prediction of soil organic carbon and total nitrogen affected by mine using Vis–NIR spectroscopy coupled with machine learning algorithms in calcareous soils

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Nov. 14, 2024

The utilization of visible-near infrared (Vis–NIR) spectroscopy presents a nondestructive, fast, reliable and cost-effective approach to predicting total nitrogen (TN) organic carbon (OC) levels. This study employed combination Vis–NIR spectroscopy, partial least-squares regression (PLSR), support vector machine (SVM) models investigate the effects mining on TN OC stocks in both topsoil (0–10 cm) subsoil (10–40 cm). 105 soil samples were collected from agricultural areas near an iron mine, polluted, moderately-polluted, non-polluted sites. Results indicated that soils at site had highest (7.5 kg m–2) (2.5 m–2), followed by moderately-polluted site. Furthermore, it was observed polluted displayed spectral reflectance. bands range 500–700 nm showed strongest correlation with content. Notably, SVM method utilizing provided superior predictions for compared PLSR methods. Additionally, demonstrated better performance (R2 = 0.87, RMSE 0.13%, RPD 2.8) 0.91, 2.4) subsoil, owing larger OM content topsoils.

Language: Английский

---

Decadal Changes of Organic Carbon, Nitrogen, and Acidity of Austrian Forest Soils

Soil Systems, Journal Year: 2022, Volume and Issue: 6(1), P. 28 - 28

Published: March 17, 2022

Repeated soil surveys provide opportunities to quantify the effect of long-term environmental change. In recent decades, topics forest acidification as a consequence acidic deposition, enrichment ecosystems with nitrogen, and loss carbon due climate change have been discussed. We used two that were 20 years apart, in order establish direction magnitude changes carbon, acidity. Soils initially sampled late 1980s. The plots revisited twenty later. Archived samples from first survey reanalyzed same protocol new samples. found stocks organic pH. However, inconsistent. general, many sites gained lost carbon. Most soils slightly enriched nitrogen. pH has not changed significantly. conclude evaluated chemical properties are mainly driven by management activities ensuing stand dynamics, atmospheric deposition. no convincing evidence effects already stock, irrespective bedrock type.

Language: Английский

---

Carbon dioxide neutrality of sustainably managed forests of Austria

Ecological Engineering, Journal Year: 2024, Volume and Issue: 209, P. 107417 - 107417

Published: Oct. 8, 2024

Language: Английский

---

Potential of contemporary earth-sheltered buildings to achieve Plus Energy status in various European climates during the heating season

Building Simulation, Journal Year: 2023, Volume and Issue: 17(1), P. 41 - 52

Published: Aug. 31, 2023

Language: Английский

---