Use Of Auxiliary Data Research Articles

Precision enhancement in variance estimation for complex environmental populations using adaptive cluster sampling

Estimating dispersion in populations that are extremely rare, hidden, geographically clustered, and hard to access is a well-known challenge. Conventional sampling approaches tend to overestimate the variance, even though it should be genuinely reduced. In this environment, adaptive cluster sampling is considered to be the most efficient sampling technique as it provides generally a lower variance than the other conventional probability sampling designs for the assessment of rare and geographically gathered population parameters like mean, total, variance, etc. The use of auxiliary data is very common to obtain the precise estimates of the estimators by taking advantage of the correlation between the survey variable and the auxiliary data. In this article, we introduced a generalized estimator for estimating the variance of populations that are rare, hidden, geographically clustered and hard-to-reached. The proposed estimator leverages both actual and transformed auxiliary data through adaptive cluster sampling. The expressions of approximate bias and mean square error of the proposed estimator are derived up to the first-order approximation using Taylor expansion. Some special cases are also obtained using the known parameters associated with the auxiliary variable. The proposed class of estimators is compared with available estimators using simulation and real data applications.

Coping with Information Loss and the Use of Auxiliary Sources of Data: A Report from the NISS Ingram Olkin Forum Series on Unplanned Clinical Trial Disruptions

While the SARS-CoV-2 (COVID-19) pandemic has led to an impressive and unprecedented initiation of clinical research, it has also led to considerable disruption of clinical trials in other disease areas, with around 80% of non-COVID-19 trials stopped or interrupted during the pandemic. In many cases the disrupted trials will not have the planned statistical power necessary to yield interpretable results. This article describes methods to compensate for the information loss arising from trial disruptions by incorporating additional information available from auxiliary data sources. The methods described include the use of auxiliary data on baseline and early outcome data available from the trial itself and frequentist and Bayesian approaches for the incorporation of information from external data sources. The methods are illustrated by application to the analysis of artificial data based on the Primary care pediatrics Learning Activity Nutrition (PLAN) study, a clinical trial assessing a diet and exercise intervention for overweight children, that was affected by the COVID-19 pandemic. We show how all of the methods proposed lead to an increase in precision relative to use of complete case data only.

Transfer Learning in Collaborative Filtering with Uncertain Ratings

To solve the sparsity problem in collaborative filtering, researchers have introduced transfer learning as a viable approach to make use of auxiliary data. Most previous transfer learning works in collaborative filtering have focused on exploiting point-wise ratings such as numerical ratings, stars, or binary ratings of likes/dislikes. However, in many real-world recommender systems, many users may be unwilling or unlikely to rate items with precision.In contrast, practitioners can turn to various non-preference data to estimate a range or rating distribution of a user's preference on an item.Such a range or rating distribution is called an uncertain rating since it represents a rating spectrum of uncertainty instead of an accurate point-wise score. In this paper, we propose an efficient transfer learning solution for collaborative filtering, known as {\em transfer by integrative factorization} (TIF), to leverage such auxiliary uncertain ratings to improve the performance of recommendation. In particular, we integrate auxiliary data of uncertain ratings as additional constraints in the target matrix factorization problem, and learn an expected rating value for each uncertain rating automatically. The advantages of our proposed approach include the efficiency and the improved effectiveness of collaborative filtering, showing that incorporating the auxiliary data of uncertain ratings can really bring a benefit. Experimental results on two movie recommendation tasks show that our TIF algorithm performs significantly better over a state-of-the-art non-transfer learning method.

Using auxiliary data to rationalize smartphone-based pre-harvest forest mensuration

Abstract Accurate mensuration of forest stands for pre-harvest planning will pose high costs if carried out by a professional forester as an on-site evaluation. The costs could be reduced if a person with limited mensuration expertise could collect the required data using a smartphone-based system such as TRESTIMA® Forest Inventory System. Without prior information, the field sample with sufficient number of measurement points over the whole stand should be selected, so that the entire variation will be covered. We present and test a rational framework based on selecting the sampling locations according to auxiliary data. As auxiliary variables, we use various spatial data sources indicating forests’ structural or spectral variation, as well as previously predicted inventory variables. We construct two variants of sampling schemes based on the local pivotal method, weighted by the auxiliary data, and compare the results to simple random sampling (SRS) with corresponding sample sizes. According to our findings, the benefits of auxiliary data depend on the considered stand, species and timber assortment. The use of auxiliary data leads generally to improved results and up to three times higher efficiency (i.e. lower variance) as compared with SRS. We conclude that the framework of applying auxiliary data has high capabilities in rationalizing the sampling efforts with little drawbacks, consequently providing potential to improve the results with similar sample size and possibility to use of non-specialists for the pre-harvest inventory.

Detection of possible hydrological precursor anomalies using long short-term memory: A case study of the 1996 Lijiang earthquake

Historical data on underground fluid can be valuable for detecting potential hydrological precursors. Although different analysis methods have been employed, nonlinear dynamic fluctuations and the lack of auxiliary observations makes hydrological precursors difficult to identify. In this study, we investigated hydrological time series of data from six monitoring sites to identify possible precursors to the Lijiang earthquake. Two long short-term memory (LSTM) models were developed to simulate and forecast hydrological variations, with and without the use of auxiliary data. The results indicate that both models are useful and valid in forecasting short-term hydrological variations. The LSTM model developed using only historical data can provide good accuracy and validity in simulating and forecasting short-term hydrological variations when no auxiliary meteorological observations are available. The results show that observations made during seismically inactive periods can be better modeled than those made during active periods. Furthermore, the exponentially weighted moving average control chart was employed as a detector to identify anomalous signals. The results show that there was a false precursor in the time series from the six monitoring points (wells and springs). Hydrological anomalies at four of the wells occurred 1–6 months before the Lijiang earthquake and can be attributed to changes in tectonic stress. However, the anomalies at one of the other wells may be unrelated to this earthquake. An anomalous signal observed at a spring could not be related to earthquakes and may have been caused by other factors. This study shows that LSTM models are useful in detecting possible hydrological precursor anomalies.

PLC-C: An Integrated Method for Sentinel-2 Topographic and Angular Normalization

Topographic and angular corrections on Sentinel-2 imagery are crucial for the generation of consistent surface reflectance. We propose a novel topographic-angular integrated normalization approach based on the combination of the path length correction (PLC) and C-factor approaches. The PLC-C normalization approach is a semiphysical method with limited use of auxiliary data: only a digital elevation model and a fixed set of kernel coefficients, ensuring its transferability for operational implementation. For the validation, we used two Sentinel-2A images over a mountainous area observed in backward (BS) and forward scattering (FS) directions from laterally adjacent orbit swaths. PLC-C significantly reduced both the topographic and directional anisotropy effects: the overlapping ratio between BS and FS observations was increased from 84.1% to 92.8% for the near-infrared band, and from 81.0% to 93.1% for the red band; the coefficient of variation of the reflectances across different aspects, which was used as a criterion of topographic effects, was reduced from 9.8%/12.2% to 3.6%/5.7% in BS/FS direction for the near-infrared band, and from 8.1%/9.7% to 4.5%/4.2% for the red band. PLC-C will contribute to the generation of analysis ready data from Sentinel-2 top of canopy reflectance.

Carbon stocks for different land cover types in Mainland Tanzania

BackgroundDeveloping countries participating in the mitigation mechanism of reducing emissions from deforestation and forest degradation (REDD+), are required to establish a forest reference emission level (FREL), if they wish to seek financial support to reduce carbon emissions from deforestation and forest degradation. However, establishment of FREL relies heavily on the accurate estimates of carbon stock as one of the input variable for computation of the emission factors (EFs). The product of an EF and activity data, such as the area of deforestation, results in the total emissions needed for establishment of FREL. This study presents the carbon stock estimates for different land cover classes based on an analysis of Tanzania’s national forest inventory data generated through the National Forest Resources Monitoring and Assessment (NAFORMA).ResultsCarbon stocks were estimated in three carbon pools, namely aboveground, belowground, and deadwood for each of the three land cover classes (i.e. Forest, non-forest, and wetland). The weighted average carbon stock was 33.35 t C ha−1 for forest land, 4.28 t ha−1 for wetland and 5.81 t ha−1 for non-forest land. The uncertainty values were 0.9% for forest land, 11.3% for wetland and 1.8% for non-forest land. Average carbon stocks for land cover sub-classes, which make up the above mentioned major land cover classes, are also presented in our study.ConclusionsThe values presented in this paper correspond to IPCC tier 3 and can be used for carbon estimation at the national scale for the respective major primary vegetation type for various purposes including REDD+. However, if local based estimates values are needed, the use of auxiliary data to enhance the precision of the area of interest is recommended.

Integrating auxiliary data in optimal spatial design for species distribution modelling

AbstractTraditional surveys used to create species distribution maps and estimate ecological relationships are expensive and time consuming. Citizen science offers a way to collect a massive amount of data at negligible cost and has been shown to be a useful supplement to traditional analyses. However, there remains a need to conduct formal surveys to firmly establish ecological relationships and trends.In this paper, we investigate the use of auxiliary (e.g. citizen science) data as a guide to designing more efficient ecological surveys. Our aim is to explore the use of opportunistic data to inform spatial survey design through a novel objective function that minimizes misclassificaton rate (i.e. false positives and false negatives) of the estimated occupancy maps. We use an initial occupancy estimate from auxiliary data as the prior in a Bayesian spatial occupancy model, and an efficient posterior approximation that accounts for spatial dependence, covariate effects, and imperfect detection in an exchange algorithm to search for the optimal set of sampling locations to minimize misclassification rate.We examine the optimal design as a function of the detection rate and quality of the citizen‐science data, and compare this optimal design with several common ad hoc designs via an extensive simulation study. We then apply our method to eBird data for the brown‐headed nuthatch in the Southeast US.We argue that planning a survey with the use of auxiliary data improves estimation accuracy and may significantly reduce the costs of sampling.

Long short-term memory neural network for air pollutant concentration predictions: Method development and evaluation

Air pollutant concentration forecasting is an effective method of protecting public health by providing an early warning against harmful air pollutants. However, existing methods of air pollutant concentration prediction fail to effectively model long-term dependencies, and most neglect spatial correlations. In this paper, a novel long short-term memory neural network extended (LSTME) model that inherently considers spatiotemporal correlations is proposed for air pollutant concentration prediction. Long short-term memory (LSTM) layers were used to automatically extract inherent useful features from historical air pollutant data, and auxiliary data, including meteorological data and time stamp data, were merged into the proposed model to enhance the performance. Hourly PM2.5 (particulate matter with an aerodynamic diameter less than or equal to 2.5 μm) concentration data collected at 12 air quality monitoring stations in Beijing City from Jan/01/2014 to May/28/2016 were used to validate the effectiveness of the proposed LSTME model. Experiments were performed using the spatiotemporal deep learning (STDL) model, the time delay neural network (TDNN) model, the autoregressive moving average (ARMA) model, the support vector regression (SVR) model, and the traditional LSTM NN model, and a comparison of the results demonstrated that the LSTME model is superior to the other statistics-based models. Additionally, the use of auxiliary data improved model performance. For the one-hour prediction tasks, the proposed model performed well and exhibited a mean absolute percentage error (MAPE) of 11.93%. In addition, we conducted multiscale predictions over different time spans and achieved satisfactory performance, even for 13–24 h prediction tasks (MAPE = 31.47%).

Use of auxiliary data of topography, snow and ice to improve model performance in a glacier-dominated catchment in Central Asia

Whether coupling auxiliary information (except for conventional rainfall–runoff and temperature data) into hydrological models can improve model performance and transferability is still an open question. In this study, we chose a glacier catchment to test the effect of auxiliary information, i.e., distributed forcing input, topography, snow-ice accumulation and melting on model calibration–validation and transferability. First, we applied the point observed precipitation and temperature as forcing data, to test the model performance in calibration–validation and transferability. Second, we took spatial distribution of forcing data into account, and did the same test. Third, the aspect was involved to do an identical experiment. Finally, the snow–ice simulation was used as part of the objective function in calibration, and to conduct the same experiment. Through stepwisely accounting these three pieces of auxiliary information, we found that a model without involving forcing data distribution, local relief, or snow–ice data can also perform well in calibration, but adding forcing data distribution and topography can dramatically increase model validation and transferability. It is also remarkable that including the snow–ice simulation into objective function did not improve model performance and transferability in this study. This may be because the well-gauged hydro-meteorological data are sufficient to constrain a well-designed hydrological model.

Comparing data mining classifiers to predict spatial distribution of USDA-family soil groups in Baneh region, Iran

Digital soil mapping involves the use of auxiliary data to assist in the mapping of soil classes. In this research, we investigate the predictive power of 6 data mining classifiers, namely Logistic regression (LR), artificial neural network (ANN), support vector machine (SVM), K-nearest neighbour (KNN), random forest (RF), and decision tree model (DTM) to create a DSM across an area covering of 3000ha in Kurdistan Province, north-west Iran. In this area, using the conditioned Latin hypercube sampling method, 217 soil profiles were selected, sampled, analysed and allocated to taxonomic classes according to Soil Taxonomy up to family level. To test the user accuracy (UA) we established a calibration and validation set (70:30%). Of the 5 soil family classes we map, the highest overall accuracy (0.71) and kappa index (0.69) are achieved using the DTA and ANN method. More specifically, the UA of prediction was up to 18.33% better in comparison to LR. Moreover, our results showed that no improvement was obtained in prediction accuracy of DTA algorithm with minimizing taxonomic distance compared to minimizing misclassification error (0.71). Overall, our results suggest that the developed methodology could be used to predict soil classes in the other regions of Iran.

Building Extraction from Airborne Laser Scanning Data: An Analysis of the State of the Art

This article provides an overview of building extraction approaches applied to Airborne Laser Scanning (ALS) data by examining elements used in original publications, such as data set area, accuracy measures, reference data for accuracy assessment, and the use of auxiliary data. We succinctly analyzed the most cited publication for each year between 1998 and 2014, resulting in 54 ISI-indexed articles and 14 non-ISI indexed publications. Based on this, we position some built-in features of ALS to create a comprehensive picture of the state of the art and the progress through the years. Our analyses revealed trends and remaining challenges that impact the community. The results show remaining deficiencies, such as inconsistent accuracy assessment measures, limitations of independent reference data sources for accuracy assessment, relatively few documented applications of the methods to wide area data sets, and the lack of transferability studies and measures. Finally, we predict some future trends and identify some gaps which existing approaches may not exhaustively cover. Despite these deficiencies, this comprehensive literature analysis demonstrates that ALS data is certainly a valuable source of spatial information for building extraction. When taking into account the short civilian history of ALS one can conclude that ALS has become well established in the scientific community and seems to become indispensable in many application fields.

Event Data on Armed Conflict and Security: New Perspectives, Old Challenges, and Some Solutions

This article presents the Event Data on Conflict and Security (EDACS) dataset, discusses the inherent problems of georeferenced conflict data, and shows how these challenges are met within EDACS. Based on an event data approach, EDACS contributes to the growing number of novel georeferenced datasets that allow researchers to identify causal pathways of violence and the dynamics of (transboundary) violence through spatiotemporal disaggregation. However, the unreflected use of any of these datasets will give researchers unjustified confidence in their findings, as the pitfalls are many and propagating errors can result in misleading conclusions. To identify and handle the different challenges to overall event data quality, we argue in favor of transparency in the data collection and coding process, to empower analysts to challenge the data and avoid cascading errors. In particular, we investigate how the choice of news sources, the handling of geographic precision, and the use of auxiliary data can bias event data. We demonstrate how the EDACS dataset design enables the analyst to deal with these issues by providing a set of variables indicating the news sources, possible sources of bias, and detailed information on geographic precision. This allows for a flexible use of the data based on individual analytical requirements.

Spatial capture‐recapture models for search‐encounter data

Summary1. Spatial capture–recapture models make use of auxiliary data on capture location to provide density estimates for animal populations. Previously, models have been developed primarily for fixed trap arrays which define the observable locations of individuals by a set of discrete points.2. Here, we develop a class of models for ‘search‐encounter’ data, i.e. for detections of recognizable individuals in continuous space, not restricted to trap locations. In our hierarchical model, detection probability is related to the average distance between individual location and the survey path. The locations are allowed to change over time owing to movements of individuals, and individual locations are related formally by a model describing individual activity or home range centre which is itself regarded as a latent variable in the model. We provide a Bayesian analysis of the model in WinBUGS, and develop a custom MCMC algorithm in the R language.3. The model is applied to simulated data and to territory mapping data for the Willow Tit from the Swiss Breeding Bird Survey MHB. While the observed density was 15 territories per nominal 1 km2plot of unknown effective sample area, the model produced a density estimate of 21·12 territories per square km (95% posterior interval: 17–26).4. Spatial capture–recapture models are relevant to virtually all animal population studies that seek to estimate population size or density, yet existing models have been proposed mainly for conventional sampling using arrays of traps. Our model for search‐encounter data, where the spatial pattern of searching can be arbitrary and may change over occasions, greatly expands the scope and utility of spatial capture–recapture models.

Improving model structure and reducing parameter uncertainty in conceptual water balance models through the use of auxiliary data

The use of uncertainty analysis is gaining considerable attention in catchment hydrological modeling. In particular, the choice of appropriate model structure, identifiability of parameter values, and the reduction of model predictive uncertainty are deemed as essential elements of hydrological modeling. The chosen model structure must be parsimonious, and the parameters used must either be derivable from field‐measured data or inferred unambiguously from analysis of catchment response data. In this paper, a long‐term water balance model for the Susannah Brook catchment in Western Australia has been pursued using the “downward approach,” which is a systematic approach to determine the model with the minimum level of complexity, with parameter values that, in theory, are derivable from existing physiographic data relating to the catchment. Through analysis of rainfall‐runoff response at three different timescales and exploring the climate, soil, and vegetation controls on the water balance response at these timescales, an initial model structure was formulated, and a priori model parameter values were estimated. Further investigation with the use of auxiliary data such as deuterium concentration in the streamflow exposed inadequacies in the chosen model structure. Two more model structures were then proposed and investigated through formulating alternative hypotheses regarding the underlying causes of observed variability, including those suggested by observed deuterium composition in the streamflows and observed groundwater level dynamics. Along the way, the resulting models were systematically evaluated in three dimensions: ability to reproduce observations, predictive uncertainty, and physical realism. The final model, which included an efficient but detailed representation of unsaturated zone time delay, was found to be superior on all three counts, i.e., improved performance, reduced predictive uncertainty, and improved physical realism. These improvements can be directly attributed to the use of the auxiliary data (deuterium composition and groundwater level dynamics), which identified weaknesses in previous model structures, and the multiple wetting front model of water movement in the unsaturated zone, which captured more effectively the time delay in the unsaturated zone.

Use of auxiliary data in semi-parametric spatial regression with nonignorable missing responses

We propose a method for reducing the error of the prediction of a quantity of interest when the outcome has missing values that are suspected to be nonignorable and the data are correlated in space. We develop a maximum likelihood approach for the parameter estimation of semi-parametric regressions in a mixed model framework. We apply the proposed method to phytoplankton data collected at fixed stations in the Chesapeake Bay, for which chlorophyll data coming from remote sensing are available. A simulation study is also performed. The availability of a variable correlated to the response allows us to achieve a substantial reduction of the prediction error of the expected value of the smoother, without having to specify a nonignorable model.

Spatial distribution of Galactic H ii regions

We present a new, detailed, analysis of the spatial distribution of Galactic H ii regions, exploiting a far richer data base than used in previous analyses. Galactocentric distances have been derived for 550 objects. Distances from the Sun could be unambiguously derived from velocity data for 117 of them, lying either outside the solar circle (84) or on a line of sight tangential to their orbit (33). For 177 further sources, distance estimates are made possible by the use of auxiliary data. A highly significant correlation between luminosity and linear diameter was found and the corresponding least-squares linear relationship in the log–log plane was used to resolve the distance ambiguity for an additional 256 sources. Within the solar circle the thickness of the distribution of H ii regions around the Galactic plane was found to be comparable to that of OB stars. At larger galactocentric radii the shape of the distribution reflects that of the warp, and its thickness along the z axis increases with increasing distance from the Galactic centre. We also confirm, for a much larger sample, the previously reported positive gradient of electron temperature with galactocentric distance.

Use of auxiliary data to improve the precision of estimators of thematic map accuracy

Collecting reference samples for accuracy assessment is expensive, so that statistical procedures that improve the precision (reduce variance) of estimators of thematic map accuracy while incurring little additional cost are desirable. Auxiliary data available from existing landcover maps, aerial photographs, videography, or AVHRR, for example, can be used in combination with ground visits to enhance the precision of map accuracy estimates. This precision gain is achieved by employing a regression estimator commonly used in survey sampling. Practically important gains in precision can be achieved with no additional complication in the sampling design and without additional field visits. The only additional cost incurred is that of obtaining the auxiliary data. The regression estimator can be used with simple random or systematic sampling if auxiliary information is available on the entire region represented by the land cover map. Alternatively, a more practical design for accuracy assessment would be double sampling, which requires that the auxiliary information be obtained for only a subsample of the map region. Appropriate estimators and standard errors are presented for using the regression estimator with either simple random or double sampling.

On the use of auxiliary data to estimate the survival function and its variance: An application to acquired immunodeficiency syndrome

Prospective studies have as their goal the estimation of the survival function when the time to the specified event may be censored due to loss to follow-up or to the termination of the study before the event of interest occurs. In such a study, information about an auxiliary event, correlated to the event of interest, is often available. An example of such an auxiliary event in cancer studies is remission or relapse. A stochastic model was proposed by Lagakos, which utilizes this type of information in the analysis of survival studies [1]. The primary objective of using an auxiliary information is to improve the estimation of survival. This article proposes a method to estimate the variance of the estimator of the survival function S( t) for the model including such auxiliary information. Thus, we compute for different situations the relative efficiency of the estimator of S( t) using the stochastic model to the estimator of S(t) using only survival data. The method is applied to data from a prospective study of 379 HIV-seropositive homosexual men, of whom 31 developed AIDS. In our example, the auxiliary event is defined by the level of CD4 lymphocyte counts using distinct threshold values, for instance 200 cells/mm 3, while the event of interest is the time to development of AIDS.

SPOT data in the digital mapping of Danish forests

Satellite remote sensing has long been understood as a potentially efficient tool for forest management planning. In order to accelerate the integration of these two techniques, a major Nordic research project was launched in 1986. Research teams from Denmark, Finland, Norway and Sweden cooperated in the project which was coordinated by the Swedish University of Agricultural Sciences and financially supported by Samarbetsnämn‐den for Nordisk Skogsforskning. The results of the research project are being published as four separate articles in the Scandinavian Journal of Forest Research. In this issue, the work done at the Bureau of Land Data and the Forest and Nature Agency in Denmark is reported. Two images from the SPOT‐1 satellite have been analyzed. The separability of forest tree species in closed stands is investigated. A new classification method is presented, and test classifications with special emphasis on forest stands are made, using varying level of thematic aggregation. Mapping accuracies between 90 and 99% were found. Regeneration areas presented classification problems, and could not be mapped reliably. The use of auxiliary data, such as existing forest boundaries, may alleviate this problem. Preliminary tests on the use of image texture as auxiliary information showed promising results. Transfer of boundary lines from classification results to a GIS system was tested, and clarified the need for integration of image‐ and digital map data, in order to utilize the image data most efficiently in forest management planning.