Yin-Chung Huang’s scientific contributions

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (7)


Figure 4: Distribution of prediction interval width of Monte Carlo dropout, conformal prediction, and Monte Carloconformal prediction for in-domain and out-of-domain samples.
Example dataset of conformal prediction containing 100 samples. The nonconformity scores are ranked from minimum to maximum.
Architecture of the convolutional neural network. ReLU stands for rectified linear unit.
Using Monte Carlo conformal prediction to evaluate the uncertainty of deep learning soil spectral models
  • Preprint
  • File available

January 2025

·

112 Reads

Yin-Chung Huang

·

·

·

Uncertainty quantification is a crucial step for the practical application of soil spectral models, particularly in supporting real-world decision making and risk assessment. While machine learning has made remarkable strides in predicting various physiochemical properties of soils using spectroscopy, predictions devoid of quantified uncertainty offer limited utility in guiding critical decisions. However, uncertainty quantification remains underutilised in the reporting of soil spectral models, with existing methods facing significant limitations. These approaches are either computationally demanding, fail to achieve the desired coverage of observed data, or struggle to handle out-of-domain uncertainty effectively. This study introduces the innovative use of Monte Carlo conformal prediction (MC-CP) as a novel approach to quantify uncertainty in the prediction of clay content from mid-infrared spectroscopy. We compared MC-CP with two established methods: (1) Monte Carlo dropout and (2) conformal prediction. Monte Carlo dropout generates prediction intervals for each sample and is effective at addressing larger uncertainties associated with out-of-domain data. However, it falls short in achieving the desired coverage – its 90 % prediction intervals only covered the observed values in 74 % of cases, well below the expected 90 % coverage. Conformal prediction, on the other hand, guarantees ideal coverage of true values but generates unnecessarily wide prediction intervals, making it overly conservative for many practical applications. In contrast, MC-CP successfully combines the strengths of both methods. It achieved a prediction interval coverage probability of 91 %, closely matching the expected 90 % coverage, and far surpassing the performance of Monte Carlo dropout. Additionally, the mean prediction interval width for MC-CP was 9.05 %, narrower than conformal prediction’s 11.11 %, while still effectively addressing the higher uncertainty in out-of-domain samples. By generating accurate prediction intervals alongside point predictions, MC-CP demonstrated its ability to deliver practical and reliable uncertainty quantification. This breakthrough enhances the real-world applicability of soil spectral models and represents a significant advancement in the field of soil science. The success of MC-CP paves the way for its integration into large-scale machine-learning models, such as soil inference systems, further revolutionising decision-making and risk assessment in soil science.

Download

Soil Science-Informed Machine Learning

November 2024

·

747 Reads

·

5 Citations

Geoderma

·

·

·

[...]

·

Machine learning (ML) applications in soil science have significantly increased over the past two decades, reflecting a growing trend towards data-driven research addressing soil security. This extensive application has mainly focused on enhancing predictions of soil properties, particularly soil organic carbon, and improving the accuracy of digital soil mapping (DSM). Despite these advancements, the application of ML in soil science faces challenges related to data scarcity and the interpretability of ML models. There is a need for a shift towards Soil Science-Informed ML (SoilML) models that use the power of ML but also incorporate soil science knowledge in the training process to make predictions more reliable and generalisable. This paper proposes methodologies for embedding ML models with soil science knowledge to overcome current limitations. Incorporating soil science knowledge into ML models involves using observational priors to enhance training datasets, designing model structures which reflect soil science principles, and supervising model training with soil science-informed loss functions. The informed loss functions include observational constraints, coherency rules such as regularisation to avoid overfitting, and prior or soil-knowledge constraints that incorporate existing information about the parameters or outputs. By way of illustration, we present examples from four fields: digital soil mapping, soil spectroscopy, pedotransfer functions, and dynamic soil property models. We discuss the potential to integrate process-based models for improved prediction, the use of physics-informed neural networks, limitations, and the issue of overparametrisation. These approaches improve the relevance of ML predictions in soil science and enhance the models’ ability to generalise across different scenarios while maintaining soil science principles, transparency and reliability.




Differentiation of fine-textured podzolic soils controlled by climate and landscape in Taiwan

December 2022

·

34 Reads

·

2 Citations

Geoderma

Podzolization is considered to be facilitated by factors such as cool and humid climate, sufficient supply of certain types of organic acids, and coarse-textured parent materials. However, podzolic soils in subalpine areas in Taiwan are characterized by clay content higher than 30%, which differed from that in typical podzolic soils in temperate and boreal regions. This study examined the importance of factors controlling the differentiation of fine-textured podzolic soils in Taiwan. Samples from 31 pedons were examined in this study, and the pedons were grouped into four types, namely Spodosols, Ultisols, Inceptisols, and Inceptisols with placic horizons. Soil colors with 7.5YR 5/6 (the accumulation of spodic materials) and micromorphological observation of oriented clay coatings in the Spodosols confirmed the occurrence of podzolization and clay illuviation, and the selective extraction of Fe and Al confirmed that the Spodosols (Type I) had a higher accumulation of organometallic complexes than the other podzolic soils. A two-step mechanism, namely (1) clay illuviation at an early stage of pedogenesis and (2) podzolization after the canopy vegetation formed, was proposed for the formation of fine-textured Spodosols. Photomicrographs of the Ultisols (Type II) displayed clay hypocoatings masked by spodic materials, and argillic horizons were characterized by a high amount of nanocrystalline iron and high optical density of the oxalate extract. However, the accumulated nanocrystalline iron oxides in the Ultisols were due to the degradation of organic matter under high summer temperatures. Moreover, the leaching of iron in the albic horizons of the Ultisols was attributed to episaturation caused by high clay contents. The Inceptisols (Type III) were transitional soils before forming the Spodosols or the Ultisols, and high clay contents further induced episaturation and formed placic horizons in the Inceptisols (Type IV). Principal component analysis revealed that soil texture and precipitation were the main factors controlling the differentiation of podzolic soils in Taiwan. These factors determined the penetrability and leaching potential of soils, and only the coarse-textured soils with higher leaching potential had the chance of developing Spodosols. However, the developments of Ultisols and Inceptisols with placic horizons were controlled by air temperature and landscape, namely elevation and slope. The bleached E horizons in podzolic Ultisols were attributed to their flat landscape, which favored episaturation and caused the reduction of iron. Low elevation and high temperature in these regions also facilitated the degradation of organic matter and the accumulation of nanominerals.



Silicon availability in relation to soil properties in Inceptisols on uncultivated lands and paddy fields in Taiwan

June 2021

·

48 Reads

·

7 Citations

Geoderma Regional

Silicon (Si), which mainly exists in soils as orthosilicic acid (H4SiO4), is beneficial for rice growth. In total, 17 Inceptisols soil profiles, 4 from uncultivated lands and 13 from paddy fields derived from felsic to ultramafic parent materials were collected for the extraction of available Si from horizon soil samples by using an acetate buffer solution. The aim of this work was to illustrate the relationship between extractable Si and general soil properties for predicting the availability of Si for rice growth. The means of available Si in the studied Inceptisols on the uncultivated lands and paddy fields were 244 mg kg⁻¹ and 114 mg kg⁻¹, respectively. The amount of available Si increased from the surface to subsurface soils in the soils on the uncultivated lands, whereas the available Si was irregular along with depth in the paddy soils. The amount of available Si was linearly and significantly (p < 0.001) correlated with clay and dithionite-citrate-bicarbonate extractable iron oxides (Fed) in all soils. The amount of available Si increased with an increase in pH in the soils from the uncultivated lands. However, a poor correlation was observed between pH and available Si in the paddy soils because labile Si was affected by human activities, such as liming and flooding, especially for surface soils. Nevertheless, the available Si was successfully estimated using crystalline iron oxides in the soils on the uncultivated lands; however, in the paddy soils, clay and organic carbon were used in the multi-regression model to estimate the amount of available Si.

Citations (4)


... Lastly, monitoring is essential in the CE of soils to ensure sustainable and efficient management of soil resources. To maintain soil functionality, monitoring can track soil health indicators like organic carbon, nutrient levels, soil aggregate stability, and microbial activities (Ma et al., 2019;Minasny et al., 2024;Padarian et al., 2022). Monitoring ensures the effective recycling of nutrients and organic matter by evaluating the efficiency of resource use, such as biochar, compost, and other amendments. ...

Reference:

Conceptualizing core aspects of circular economy in soil: A critical review and analysis
Soil Science-Informed Machine Learning

Geoderma

... De Oliveira et al. (2023) used artificial neural networks, SVM, random forest (RF), and gradient boosting to compare and enhance the accuracy of soil classifications. For soil horizon level classification, Fajardo et al. (2016) introduced the fuzzy clustering technique, and Huang et al. (2023) applied PCA and k-means clustering to demonstrate the possibility of achieving precision in soil classification. Only a few studies used MIR spectroscopy (Xu et al., 2020;Zhang et al., 2021). ...

Using pXRF and Vis-NIR for characterizing diagnostic horizons of fine-textured podzolic soils in subtropical forests

Geoderma

... The metallic composition of spodic material can be detected by pXRF (Weindorf et al., 2012), and both OC and sesquioxides show distinct features in Vis-NIR reflectance (Demattê et al., 2014). Moreover, Huang et al. (2022) summarized the podzolization processes in Taiwan and found clay illuviation also coincides with podzolization to form fine-textured Spodosols. Hence, the fine-textured podzolic soils provide horizons with diverse concentrations of OC, sesquioxides, and clay through the profile. ...

Differentiation of fine-textured podzolic soils controlled by climate and landscape in Taiwan
  • Citing Article
  • December 2022

Geoderma

... Factors affecting crop-available silicon and the estimation of PhytOC (phytolithoccluded carbon) involve a combination of environmental, soil-related, and plantspecific factors (Matichenkov and Bocharnikova 2001;Huang and Hseu 2021;Anggria et al. 2020). ...

Silicon availability in relation to soil properties in Inceptisols on uncultivated lands and paddy fields in Taiwan
  • Citing Article
  • June 2021

Geoderma Regional