Friday, November 20, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News

Two Stage Iterative Method Restores Incomplete Small Scale River Water Quality Datasets

November 20, 2026
in Science News
Margaret Porter
By Margaret Porter Scienmag Editorial Profile - Biodiversity Science
Reading Time: 7 mins read
0
Two Stage Iterative Method Restores Incomplete Small Scale River Water Quality Datasets

Two Stage Iterative Method Restores Incomplete Small Scale River Water Quality Datasets

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

Environmental monitoring networks frequently struggle with discontinuous data records caused by equipment failures, routine maintenance, natural hazards, and transmission interruptions, yet maintaining continuous observations remains critical for effective aquatic management. In a study published in the journal Marine Development, researchers Fangyu Wang, Xiaoying Cui, Yuying Gui, and Yanxin Qiao introduced and validated a two-stage iterative imputation framework specifically engineered to reconstruct missing observations in small-scale river monitoring archives. Real-time water quality assessments depend heavily on unbroken data streams to track pollutant transport and verify ecosystem stability, but monitoring networks often accumulate gaps that distort statistical calculations and limit analytical accuracy. When datasets are constrained in size, traditional single-pass algorithms or complex machine learning models frequently falter because small sample volumes fail to supply adequate training examples. To overcome these limitations, the research team structured their strategy around empirical evaluations of multiple filling algorithms across different missingness patterns, proving that a targeted stepwise protocol can systematically restore incomplete time series while respecting the underlying physical and temporal characteristics of river monitoring metrics.

The foundational challenge identified by the investigators stems from the common practice of omitting incomplete records, an approach that directly shrinks sample sizes and sacrifices crucial information regarding watershed dynamics. In standard hydrological analyses, deleting entire records that contain missing entries in one or more features inevitably reduces sample utility, creates uneven data matrices, and introduces statistical bias. Although basic substitution procedures such as inserting the mean, median, or mode are frequently employed due to their computational simplicity, the authors emphasize that these static methods fail to account for internal correlations across time or among related environmental variables. Similarly, while advanced architectures including artificial neural networks, generative adversarial networks, and multilayer perceptrons demonstrate strong capabilities when processing extensive data repositories, their heavy parameter demands and rigorous training requirements render them unsuitable for compact, small-scale environmental records. Furthermore, univariate statistical alternatives like linear interpolation connect neighboring endpoints but cannot incorporate external influences, whereas standard time-series models require substantial historical depth and can struggle to maintain forward coherence when forecasting prolonged interruptions across sparse historical records.

To rigorously test their methodology, the authors examined subdaily water quality measurements collected over an approximate two-year duration spanning from July 1, 2018, to June 30, 2020, by the Environmental Protection Department of Shandong province. The field measurements originated from the Fanli station positioned within the Xiaoqing River basin in China, a critical land-source waterway that originates in Jinan City and represents the primary source of pollution entering Laizhou Bay. Sustaining the ecological equilibrium of Laizhou Bay requires precise tracking of upstream discharges, making the water quality of the Xiaoqing River an essential determinant of marine conservation and sustainable resource utilization. Within this monitoring context, the researchers focused on three primary target indicators: the chemical oxygen demand permanganate index, total nitrogen concentration, and total phosphorus concentration. These specific chemical variables serve as supportive water quality indicators vital for environmental analysis, regulatory control, and broader aquatic habitat protection, even though automated field stations regularly face unexpected communication dropouts, seasonal disruptions, and sensor servicing that leave unrecorded intervals across successive monitoring cycles.

Preliminary exploratory analysis revealed that the raw chemical observations contained notable variations, realistic statistical outliers, and differing patterns of temporal discontinuity across the three monitoring parameters. When reviewing the raw distributions through box plots, the researchers chose to retain all identified statistical outliers exceeding 1.5 times the interquartile range because these values reflected authentic physical occurrences rather than erroneous sensor artifacts. Among the three examined parameters, total phosphorus exhibited greater baseline stability across the observation window compared to the broader numerical fluctuations recorded for total nitrogen and chemical oxygen demand. Missingness distribution mapping further demonstrated that missing data rates remained relatively consistent within individual years, yet missing timestamps did not display synchronized or uniform patterns across the three chemical indicators. Given that the underlying causes of absent readings involved periodic maintenance routines and occasional transmission dropouts rather than systemic physical exclusions, the missing values were reasonably classified under the missing at random mechanism, providing a valid theoretical foundation for applying statistical and machine learning imputation models.

A critical step in designing the restorative methodology involved quantifying the duration of observation gaps, which directly dictated the separation between the initial and subsequent computational stages. Frequency analysis showed that approximately one-third of all missing intervals lasted only four hours, which corresponds to the standard recording interval between consecutive automated measurements and indicates that the majority of missing entries occurred through brief, isolated incidents. Conversely, gaps extending beyond 336 hours, representing continuous missing durations of two weeks or more, comprised more than four percent of the recorded data and typically stemmed from substantial equipment breakdowns requiring lengthy mechanical servicing. Recognizing that mathematical imputation accuracy generally deteriorates as missing intervals widen, the investigators established 336 hours as the structural boundary separating their computational pipeline. This quantitative threshold designated interruptions shorter than 336 hours as small-range random omissions and gaps equal to or exceeding 336 hours as large-scale continuous interruptions, allowing more than 95 percent of the dataset to be stabilized during the initial phase before tackling the remaining long-term gaps.

To assist multivariate imputation procedures without distorting natural environmental relationships, the investigators examined potential dependencies by computing nonparametric Spearman rank correlation coefficients across fully observed, normalized data segments. Incorporating auxiliary water quality parameters, often designated as helper variables, enables multivariate models to infer missing chemical concentrations from synchronous changes observed across companion sensors. The correlation analysis identified distinct interdependencies among the measured variables, revealing that chemical oxygen demand correlated notably with water temperature, pH, and electrical conductivity measurements. Concurrently, total nitrogen and total phosphorus demonstrated significant correlations with water temperature, pH, dissolved oxygen, and turbidity. Despite noting that the helper variables themselves suffered from minor percentages of missing data, the researchers retained all raw auxiliary records without artificial filtering or supplementary modifications. Preserving the original state of these supporting records ensured that subsequent multivariate algorithms operated under realistic operational monitoring constraints rather than idealized laboratory assumptions.

During the first phase of the experimental framework, the research team evaluated multiple imputation techniques on short-term random gaps using continuous observation series exceeding 336 hours, corresponding to at least 84 sequential measurement points. Artificial missingness scenarios were generated by masking data points across five designed rates of 5%, 10%, 20%, 30%, and 40%, with each missing condition replicated 10,000 times per target variable to ensure statistical rigor. Because the irregular distribution of randomized gaps rendered the autoregressive integrated moving average model unsuitable for this stage, the team compared mean imputation, median imputation, linear interpolation, k-nearest neighbors, random forest, and multiple imputation by chained equations. Across all evaluated metrics—including root mean square error, mean absolute error, mean absolute percentage error, Nash–Sutcliffe efficiency, and the coefficient of determination—linear interpolation consistently delivered superior performance, followed by k-nearest neighbors and random forest. For linear interpolation, the average mean absolute percentage error remained below 10% across all three chemical variables, aligning with standard instrument measurement uncertainty and justifying its deployment to populate short-term gaps within the primary dataset.

With short-term random intervals successfully filled by linear interpolation, the augmented records and their corresponding helper covariates were advanced into the second stage to resolve the remaining long-term continuous gaps. This secondary phase addressed the prolonged missing sequences that accounted for less than 5% of the total dataset, where preserving longitudinal time-series coherence became paramount. To evaluate model behavior under extended gap conditions, the final 85 consecutive observations of each series were masked as a dedicated testing set while preceding records served as training inputs for the autoregressive integrated moving average model alongside the alternative algorithms. Comparative performance assessments demonstrated that no single imputation algorithm universally dominated across all water quality parameters during prolonged interruptions. Instead, optimal performance depended heavily on parameter-specific variability and covariate completeness: the autoregressive integrated moving average model proved optimal for chemical oxygen demand, multiple imputation by chained equations achieved the greatest accuracy for total nitrogen, and linear interpolation performed best for total phosphorus.

Quantitative evaluations confirmed that assigning customized, parameter-specific algorithms during the second stage maintained strong statistical congruence between original and imputed observations across all chemical variables. When analyzing regression relationships between original values and model predictions, all three selected methods maintained high Nash–Sutcliffe efficiency coefficients exceeding 0.89, providing solid empirical confirmation of their restorative reliability. Although scatterplots indicated that minor deviations persisted between estimates and actual observations, boxplot comparisons demonstrated that the distribution means of the imputed values matched the original metrics closely while exhibiting fewer extreme outliers. The resulting mean absolute percentage error values for the secondary phase reached 12.58% for chemical oxygen demand using the autoregressive integrated moving average model, 7.54% for total nitrogen using multiple imputation by chained equations, and 17.76% for total phosphorus using linear interpolation. By systematically applying these specialized techniques to their corresponding chemical series, the two-stage iterative architecture successfully generated fully reconstructed, continuous monitoring profiles across the entire two-year study period.

An empirical assessment of the fully reconstructed time series at the Fanli monitoring station demonstrated that the two-stage iterative framework preserved essential hydrological dynamics while confirming a broad environmental recovery trend. Over the monitored duration, chemical oxygen demand, total nitrogen, and total phosphorus concentrations displayed an overall downward trajectory, demonstrating the practical efficacy of water resource management and pollution mitigation initiatives within the Xiaoqing River basin. Although the authors observed that extended continuous gap imputations occasionally produced subdued fluctuations or direct linear segments due to incomplete helper covariates, the staged protocol avoided the severe data attrition associated with record deletion. By maximizing information extraction from limited initial records and systematically bridging the gap between brief fluctuations and long-term sensor failures, the researchers established a reliable mathematical roadmap for water quality monitoring, offering a practical reference framework that can be adapted to other data-sparse environmental monitoring networks.

Subject of Research: Two-stage iterative imputation of missing values in small-scale river water quality datasets

Article Title: Two stage iterative approach for addressing missing values in small-scale water quality data

Article References: Wang, F., Cui, X., Gui, Y., & Qiao, Y. (2024). Two stage iterative approach for addressing missing values in small-scale water quality data. Marine Development, 2(1), Article 27. https://doi.org/10.1007/s44312-024-00040-3

Image Credits: AI Generated

DOI: 10.1007/s44312-024-00040-3

Keywords: water quality monitoring, missing data imputation, Xiaoqing River, Fanli station, linear interpolation, autoregressive integrated moving average, multiple imputation by chained equations, random forest imputation, k-nearest neighbor, chemical oxygen demand, total nitrogen, total phosphorus, small-scale datasets, missing at random, environmental data analysis, hydrological time series, mean absolute percentage error

Cite Scienmag News

Margaret Porter. (November 20, 2026). Two Stage Iterative Method Restores Incomplete Small Scale River Water Quality Datasets. Scienmag. https://scienmag.com/two-stage-iterative-method-restores-incomplete-small-scale-river-water-quality-datasets/

Margaret Porter. "Two Stage Iterative Method Restores Incomplete Small Scale River Water Quality Datasets." Scienmag, 20 November 2026, https://scienmag.com/two-stage-iterative-method-restores-incomplete-small-scale-river-water-quality-datasets/. Accessed 20 November 2026.

Margaret Porter. "Two Stage Iterative Method Restores Incomplete Small Scale River Water Quality Datasets." Scienmag. November 20, 2026. https://scienmag.com/two-stage-iterative-method-restores-incomplete-small-scale-river-water-quality-datasets/

Tags: Autoregressive integrated moving averagechemical oxygen demandenvironmental data analysisFanli stationhydrological time seriesK-Nearest Neighborlinear interpolationmean absolute percentage errormissing at randomMissing data imputationmultiple imputation by chained equationsrandom forest imputationsmall-scale datasetstotal nitrogentotal phosphoruswater quality monitoringXiaoqing River
Share26Tweet16
Previous Post

Hybrid Signal Decomposition and Denoising Accurately Forecasts Weekly Nature Tourism Demand

Next Post

Full Annual Cycle Models Reveal Complex Abundance Trends for Migratory Birds

Related Posts

Bxb1 Recombinase DNA Excision Dynamically Preconfigures Bacterial Biosensor Sensitivity Regimes
Biotechnology

Bxb1 Recombinase DNA Excision Dynamically Preconfigures Bacterial Biosensor Sensitivity Regimes

November 20, 2026
Chinese Herbal Decoction Slows Triple-Negative Breast Cancer Growth in Lab Models
Biology

Chinese Herbal Decoction Slows Triple-Negative Breast Cancer Growth in Lab Models

November 20, 2026
Review Details How Age-Associated B Cells Drive Chronic Autoimmune Diseases
Medicine

Review Details How Age-Associated B Cells Drive Chronic Autoimmune Diseases

November 20, 2026
Preoperative Scopolamine Lowers Postoperative Anxiety and Depression in Gynecological Surgery Patients
Psychology & Psychiatry

Preoperative Scopolamine Lowers Postoperative Anxiety and Depression in Gynecological Surgery Patients

November 20, 2026
Essential Oil and Nisin Combinations Inactivate Food Pathogens in Model Cheese
Medicine

Essential Oil and Nisin Combinations Inactivate Food Pathogens in Model Cheese

November 20, 2026
Successive Corvis ST Tonometry Shows Minimal Corneal Biomechanical Changes in Young Myopes
Biotechnology

Successive Corvis ST Tonometry Shows Minimal Corneal Biomechanical Changes in Young Myopes

November 20, 2026
Next Post
Full Annual Cycle Models Reveal Complex Abundance Trends for Migratory Birds

Full Annual Cycle Models Reveal Complex Abundance Trends for Migratory Birds

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Bxb1 Recombinase DNA Excision Dynamically Preconfigures Bacterial Biosensor Sensitivity Regimes
  • Chinese Herbal Decoction Slows Triple-Negative Breast Cancer Growth in Lab Models
  • Review Details How Age-Associated B Cells Drive Chronic Autoimmune Diseases
  • Preoperative Scopolamine Lowers Postoperative Anxiety and Depression in Gynecological Surgery Patients

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading