Overestimation Error Research Articles

Rapid Estimation of Road Friction for Anti-Skid Autonomous Driving

The failure to estimate friction on the road is a significant cause of roadway departure crashes. It is even more relevant to autonomous vehicles (AVs) as they are currently not designed to measure the skid resistance of roads. Most traditional measurements are too inefficient for real-time AV control. Thus, in this paper, we propose a dynamic method to estimate pavement friction level using computer vision. We collect 100 sets of high-quality road images and their resistance values to train the model, and design two major methods for analysis: 1) texture identification, consisting of grayscale-enhanced local binary pattern and the gray-level gradient co-occurrence matrix and 2) a deep neural network based on domain knowledge (TLDKNet). We introduce two standards of classification for model training and propose three indices—underestimation error, overestimation error, and accuracy—to verify the performance of our algorithms. The results showed the great correlation between pavement texture and skid resistance. The TLDKNet yielded the best performance with an accuracy of 90.67% and only 2.67% underestimation error, showing that it is sufficiently conservative with regard to safety. Based on the proposed method of estimation, a framework for anti-skid driving control is developed regarding the car-following behavior and turning movements. The anti-skid control framework provides new insights into enhancing the AV safety performance.

Directional Skid Resistance Characteristics of Road Pavement: Implications for Friction Measurements by British Pendulum Tester and Dynamic Friction Tester

The British pendulum tester (BPT) and the dynamic friction tester (DFT) are commonly used to evaluate the polishing effect of traffic on pavement skid resistance. The present study was conducted to answer two questions: (i) whether the directional moving traffic action results in significant directional differences in skid resistance of the pavement, and (ii) if so, whether BPT and DFT are effective tools for evaluating the skid resistance performance of polished pavements with significant directional skid resistance characteristics. Five in-service pavement sections with different degrees of polishing were examined in this study. The following findings were obtained from the study: (a) There were statistically significant differences between the longitudinal and transverse skid resistance of the polished pavements tested; (b) there were no statistically significant differences between the longitudinal and transverse skid resistance of the newly paved unpolished pavement tested; (c) BPT was able to differentiate the directional skid resistance differences of the polished pavements tested; and (d) DFT was unable to differentiate the directional skid resistance differences of the polished pavements tested. DFT produces one skid resistance value for a circular path at each test point. The test results found that DFT over-estimated the traffic-direction skid resistance of in-service pavements. The over-estimation errors ranged from 0.001 to 0.023 DFT units, and the corresponding percent errors were from 0.4% to 11.1%. This finding suggests that the over-estimation effect of skid resistance directionality of polished pavements on DFT measurements should be recognized in pavement skid resistance evaluation and pavement safety studies.

P.083 Radiology reporting of low-grade glioma growth underestimates tumor expansion

Background: Surveillance by serial magnetic resonance imaging (MRI) is important in the management of diffuse low-grade gliomas (LGGs). Radiological interpretations of LGG scans, however, are typically qualitative and difficult to use clinically. Methods: We retrospectively compared radiological interpretations of LGG growth/stability to volume change measured by manual segmentation. Tumour diameter was measured to evaluate methods for assessing glioma progression, including RECIST criteria, Macdonald/RANO criteria, and mean tumour diameter/ellipsoid method. Results: Tumours evaluated as stable by radiologists grew a median 5.1 mL (11.1%) relative to the comparison scan. Those evaluated as having grown increased by 13.3 mL (23.7%). Diameter-based measurements corresponded well but tended to overestimate segmented volumes, and overestimation error increased with tumour size. Agreement with segmented volume improved from a mean difference of 17.6 to 4.5 to 3.9 mm for diameter and from 104.0 to 25.3 to 15.9 mL for volume with measurements in one, two, and three dimensions. Conclusions: Given evidence that LGG volume and growth are prognostic factors, lesions should be accurately measured. Current radiological reporting workflows fail to appreciate and communicate the true expansion of LGGs. Volumetric analysis remains the gold standard for growth assessment, but diametric measurements in three dimensions may be an acceptable alternative.

Dissecting Intra-Tumor Heterogeneity by the Analysis of Copy Number Variations in Single Cells: The Neuroblastoma Case Study.

Tumors often show intra-tumor heterogeneity because of genotypic differences between all the cells that compose it and that derive from it. Recent studies have shown significant aspects of neuroblastoma heterogeneity that may affect the diagnostic-therapeutic strategy. Therefore, we developed a laboratory protocol, based on the combination of the advanced dielectrophoresis-based array technology and next-generation sequencing to identify and sort single cells individually and carry out their copy number variants analysis. The aim was to evaluate the cellular heterogeneity, avoiding overestimation or underestimation errors, due to a bulk analysis of the sample. We tested the above-mentioned protocol on two neuroblastoma cell lines, SK-N-BE(2)-C and IMR-32. The presence of several gain or loss chromosomal regions, in both cell lines, shows a high heterogeneity of the copy number variants status of the single tumor cells, even if they belong to an immortalized cell line. This finding confirms that each cell can potentially accumulate different alterations that can modulate its behavior. The laboratory protocol proposed herein provides a tool able to identify prevalent behaviors, and at the same time highlights the presence of particular clusters that deviate from them. Finally, it could be applicable to many other types of cancer.

Robust, blind multichannel image identification and restoration using stack decoder

In this study, the authors introduce new solutions and improvements to the multi-channel blind image deconvolution problem. More precisely, authors’ contributions are threefold: (i) At first, a simplified version of the existing cross-relation method for blind system identification is proposed; but most importantly, the authors incorporate into the channel estimation cost function a sparsity constraint to deal with the challenging issue of channel order overestimation errors; (ii) then, once the channel identification is achieved, a new image restoration method based on the stack decoding algorithm is introduced; and (iii) finally, a refining approach using an ‘all-at-once’ optimisation technique with an improved mixed norm regularisation is considered. The performance of the proposed approach was evaluated using several numerical simulations. Blind system identification and image restoration tasks were evaluated with respect to several criteria: numerical complexity, robustness to noise effects and channel order estimation. The results obtained are promising and highlight the effectiveness of the proposed approach.

Consistent estimation of high-dimensional factor models when the factor number is over-estimated

A high-dimensional $r$-factor model for an $n$-dimensional vector time series is characterised by the presence of a large eigengap (increasing with $n$) between the $r$-th and the $(r+1)$-th largest eigenvalues of the covariance matrix. Consequently, Principal Component (PC) analysis is the most popular estimation method for factor models and its consistency, when $r$ is correctly estimated, is well-established in the literature. However, popular factor number estimators often suffer from the lack of an obvious eigengap in empirical eigenvalues and tend to over-estimate $r$ due, for example, to the existence of non-pervasive factors affecting only a subset of the series. We show that the errors in the PC estimators resulting from the over-estimation of $r$ are non-negligible, which in turn lead to the violation of the conditions required for factor-based large covariance estimation. To remedy this, we propose new estimators of the factor model based on scaling the entries of the sample eigenvectors. We show both theoretically and numerically that the proposed estimators successfully control for the over-estimation error, and investigate their performance when applied to risk minimisation of a portfolio of financial time series.

Demographic assessment and adjustment of census age distribution in selected West African Countries

Context/background: Despite the public acknowledgment of improvement in sub-Saharan African censuses and surveys in the recent time, the quality of age-sex data still remains relatively poor. This study assessed the quality of reported age-sex data in selected West African countries. Data source and methods: We utilized the available recent West African census data to address the study objective. Data adjustments were made using the UN-moving average technique, Carrier Far-rag, Karup-king newton and Arriaga methods. Results: First, the age- distribution in five years was examined for errors. The UN-Age-sex accuracy index showed that all the selected countries had inaccurate age-distribution, with the exception of Cameroon (2005) and Senegal (2002) censuses which revealed a lesser distorted age- distribution. Utilizing the strong smoothing techniques, the results showed that virtually all the censuses were plaqued with the overestimation error at older ages, particularly the female population. The error margin ranged between 10-25%. Conclusion: The study concludes that inaccuracies in West African census data persist and recommends more quality age-sex data in region.

Evaluating the Algorithm for Correction of the Bright Band Effects in QPEs with S-, C- and X-Band Dual-Polarized Radars

The bright band, a layer of enhanced radar reflectivity associated with melting ice particles, is a major source of significant overestimation in quantitative precipitation estimation (QPE) based on the Z-R (reflectivity factor-rain rate) relationship. The effects of the bright band on radar-based QPE can be eliminated by vertical profile of reflectivity (VPR) correction. In this study, we applied bright-band correction algorithms to evaluate three different bands (S-, C- and X-band) of dual-polarized radars and to reduce overestimation errors in Z-R relationship-based QPEs. After the reflectivity was corrected by the algorithms using average VPR (AVPR) alone and a combination of average VPR and the vertical profile of the copolar correlation coefficient (AVPR+CC), the QPEs were derived. The bright-band correction and resulting QPEs were evaluated in eight precipitation events by comparing to the uncorrected reflectivity and rain-gauge observations, separately. The overestimation of Z-R relationship-based QPEs associated with the bright band was reduced after correction by the two schemes for which hourly rainfall was less than 5 mm. For the verification metrics of RMSE (root-mean-square error), RMAE (relative mean absolute error) and RMB (relative mean bias) of QPEs, averaged over all eight cases, the AVPR method improved from 2.28, 0.94 and 0.78 to 1.55, 0.60 and 0.40, respectively, while the AVPR+CC method improved to 1.44, 0.55 and 0.30, respectively. The QPEs after AVPR+CC correction had less overestimation than those after AVPR correction, and similar conclusions were drawn for all three different bands of dual-polarized radars.

SonReb concrete assessment for spatially correlated NDT data

In this paper the existence of a spatial correlation on the readings of the rebound hammer tests and ultrasonic pulse velocity tests is firstly investigated by means of experimental evaluations on two RC walls. The effect of this correlation is then transferred in the concrete strength evaluation by means of the well known SonReb method. Monte Carlo simulations, carried out to check the goodness of the evaluation with the SonReb standard procedure, highlighted a relevant probability of under- or over-estimation as well as the possibility of not negligible absolute errors of the expected mean compression strengths for the investigated structural elements. A simple modification to the standard procedure is, finally, proposed in order to take into account the concrete spatial correlation and improve the goodness of the SonReb evaluation. This modification, that basically consists in increasing the relative distance between the location of the individual test readings, allows the reduction of the unconservative maximum percentage overestimation error from roughly 15% to 5%.

Prenatal measurement of umbilical cord length using magnetic resonance imaging

ObjectiveThis study aimed to evaluate the feasibility of prenatal measurement of umbilical cord length using magnetic resonance imaging (MRI) in comparison with that of postnatal visual measurement. Study designPatients who underwent MRI pelvimetry to rule out cephalopelvic disproportion were recruited. We compared the umbilical cord length measured prenatally using the three-dimensional image of the umbilical cord constructed by AZE virtual place system (AZE®, Japan) on MRI (M-length) and measured postnatally using the actual length (R-length). The time interval between MRI scan and birth was set within 4 weeks. Patients were divided into normal, long (>90th centile), and short (<10th centile) groups based on the umbilical cord length. ResultsA total of 106 pregnancies were analyzed. MRI was performed at a mean gestational age of 37.4 (34.1–41.7) weeks; the mean gestational age at delivery was 39.4 (36.0–41.8) weeks. M-length and R-length were 67.9 ± 13.6 [mean ± standard deviation] and 57.2 ± 12.7 cm, respectively. The correlation coefficient was larger in the long and short groups than in the normal group. The intraclass correlation coefficient showed high agreement, and Pearson’s correlation coefficient revealed correlation inspection as r = 0.702. Bland–Altman analysis indicated non-agreement, and the bias and upper and lower limits of agreement were 10.6 cm and -6.0 and 27.3 cm, respectively. ConclusionsShort and long umbilical cord lengths are associated with perinatal mortality and morbidity. Prenatal measurement of umbilical cord length using MRI is feasible for daily clinical use to distinguish short or long umbilical cords. Further research is needed for more precise clinical application and technical reduction of overestimation errors.

A statistical fracture model for Antarctic ice shelves and glaciers

Abstract. Antarctica and Greenland hold enough ice to raise sea level by more than 65 m if both ice sheets were to melt completely. Predicting future ice sheet mass balance depends on our ability to model these ice sheets, which is limited by our current understanding of several key physical processes, such as iceberg calving. Large-scale ice flow models either ignore this process or represent it crudely. To model fractured zones, an important component of many calving models, continuum damage mechanics as well as linear fracture mechanics are commonly used. However, these methods have a large number of uncertainties when applied across the entire Antarctic continent because the models were typically tuned to match processes seen on particular ice shelves. Here we present an alternative, statistics-based method to model the most probable zones of the location of fractures and demonstrate our approach on all main ice shelf regions in Antarctica, including the Antarctic Peninsula. We can predict the location of observed fractures with an average success rate of 84 % for grounded ice and 61 % for floating ice and a mean overestimation error rate of 26 % and 20 %, respectively. We found that Antarctic ice shelves can be classified into groups based on the factors that control fracture location.

The IG- file use to Gauge the Apical Diameter in Endodontics: An In Vitro Study

Aim:The aim of this study was to evaluate the efficacy of the IG-file, a new instrument designed for apical diameter gauging.Materials and Methods:After shaping with F1 Universal Protaper, 60 roots were randomly divided into two groups and assigned to two operators, One Expert in Endodontics (EO) and One Unexpert (UO).In each sample, after canal curvatures have been detected, the apical diameters were measured with the IG-file and the K-NiTi. The results were compared with the reference value obtained by retrograde apical gauging. The data were statistically analyzed.Results:Among 60 samples, 10% of errors were recorded when the IG-files were used; in the K-NiTi group the incorrect measurements were 70%.In both groups (expert and unexpert) the IG-file measurements were more accurate than the K-NiTi (90 vs 33 and 90 vs 26,7). The differences were statistically significant.In curved canals, the difference between measurement rates performed with both instruments was statistically significant (85,7% IG-file vs 28,6% K-NiTi) as well as for the samples without curvatures (92,3% IG file vs 30,8% NiTi file).In root canals without curvatures overestimation errors in K-NiTi file group are more frequent than underestimation errors. This difference was statistically significant.Conclusion:A proper gauging of the apical diameter has a key role in endodontic therapy; an incorrect measurement can lead to clinical failures. This “in vitro” study highlights that IG-file improves measurement accuracy independently from clinician experience. Furthermore, in curved canals, the IG-file is more accurate than K-NiTi.

Overall success rate of a safe and efficacious drug: Results using six phase 1 designs, each followed by standard phase 2 and 3 designs.

To evaluate the overall success rate of a new drug, phase 1, 2, and 3 trials were simulated using eight toxicity and two non-decreasing efficacy profiles. Six phase 1 designs including the standard 3 + 3, CCD, BOIN, mTPI, mTPI-2, and CRM were considered with standard phase 2 and 3 designs.Based on our results, phase 1 design recommendations are provided when data informing the general shape of the dose-toxicity curve exist. If a large jump in toxicity between dose levels is expected, the standard 3 + 3 design is recommended; it more often recognized when the MTD was exceeded and had the highest overall success rates. If gradually increasing toxicity is expected, a nonstandard design other than the CRM is recommended. Nonstandard designs were more aggressive in dosing and MTD estimation than the standard 3 + 3 and had higher overall success rates, but the CRM was too aggressive and most frequently overestimated the true MTD. If fairly constant, safe toxicity is expected across dose levels, the BOIN or CRM designs are recommended; they escalated to the highest dose most frequently with superior overall success rates.Without data informing the shape of the dose-toxicity curve, nonstandard phase 1 designs with a modified excessive toxicity rule more easily eliminating unsafe dose levels are recommended. With this modification, MTD overestimation error decreased and overall success rates were similar or higher with nonstandard designs. Among nonstandard designs, the modified CCD and BOIN perform well and are as transparent and simple to implement as the standard 3 + 3 design.

Errors induced by the neglect of polarization in radiance calculations for three-dimensional cloudy atmospheres

Remote sensing instruments observe radiation being scattered and absorbed by molecules, aerosol particles, cloud droplets and ice crystals. In order to interpret and accurately model such observations, the vector radiative transfer equation needs to be solved, because scattering polarizes the initially unpolarized incoming solar radiation. A widely used approximation in radiative transfer theory is the neglect of polarization which allows to greatly simplify the radiative transfer equation.It is well known that the error caused by multiple Rayleigh scattering can be larger than 10%, depending on wavelength and sun-observer geometry (Mishchenko et al., 1994 [1]). For homogeneous plane-parallel layers of liquid cloud droplets the error is comparatively small (below 1%) (Hansen, 1971 [2]). Therefore, in radiative transfer modelling for cloud remote sensing polarization is mostly neglected. However, in reality clouds are not plane-parallel layers of water droplets but complex three-dimensional (3D) structures and observations of clouds usually include pixels consisting of clear and cloudy parts. In this study we revisit the question of the magnitude of the error due to the neglect of polarization in radiative transfer theory for a realistic 3D cloudy atmosphere.We apply the Monte Carlo radiative transfer model MYSTIC with and without neglecting polarization and compare the results. At a phase angle of 90° and 400 nm wavelength we find the maximum overestimation error of about 8% for complete clear-sky conditions. The error is reduced to about 6% in clear-sky regions surrounded by clouds due to scattering from clouds into the clear regions. Within the clouds the error is up to 4% with the highest values in cloud shadows. In backscattering direction the radiance is underestimated by about 5% in clear regions between clouds. For other sun-observer geometries, the error ranges between the two extremes. The error decreases with wavelength and in the absorption bands.

Errors in nanoparticle growth rates inferred from measurements in chemically reacting aerosol systems

Abstract. In systems in which aerosols are being formed by chemical transformations, individual particles grow due to the addition of molecular species. Efforts to improve our understanding of particle growth often focus on attempts to reconcile observed growth rates with values calculated from models. However, because it is typically not possible to measure the growth rates of individual particles in chemically reacting systems, they must be inferred from measurements of aerosol properties such as size distributions, particle number concentrations, etc. This work discusses errors in growth rates obtained using methods that are commonly employed for analyzing atmospheric data. We analyze “data” obtained by simulating the formation of aerosols in a system in which a single chemical species is formed at a constant rate, R. We show that the maximum overestimation error in measured growth rates occurs for collision-controlled nucleation in a single-component system in the absence of a preexisting aerosol, wall losses, evaporation or dilution, as this leads to the highest concentrations of nucleated particles. Those high concentrations lead to high coagulation rates that cause the nucleation mode to grow faster than would be caused by vapor condensation alone. We also show that preexisting particles, when coupled with evaporation, can significantly decrease the concentration of nucleated particles. This can lead to decreased discrepancies between measured growth rate and true growth rate by reducing coagulation among nucleated particles. However, as particle sink processes become stronger, measured growth rates can potentially be lower than true particle growth rates. We briefly discuss nucleation scenarios in which the observed growth rate approaches zero while the true growth rate does not.

An examination of microwave rainfall retrieval biases and their characteristics over the Amazon

Estimates of surface precipitation obtained from passive microwave sensors over land are closely related to the ice path present in the clouds. However, there are cloud types without any ice or with an ice layer not thick enough to justify the associated rainfall. For these cloud types, the precipitation is not estimated correctly, causing an underestimation of the precipitation. On the other hand, there are cases of deep clouds, in which the signal produced by ice scattering is not effectively associated with precipitation, producing, in turn, an overestimate of rainfall. This study analyzes cases that have large errors in the rainfall estimates obtained from passive microwave data to better understand and potentially mitigate these biases. This study uses data from the Tropical Rainfall Measuring Mission (TRMM) satellite, specifically the Precipitation Radar (PR), TRMM Microwave Imager (TMI) and Lightning Imaging Sensor (LIS). Ten years of TRMM data (2002−2011) are used in the analysis. The study area is approximately 1,110,000 km2, centered on the city of Manaus in the Amazon region. The error distribution resembled a Gaussian distribution. The error population was then divided into three categories, one class denominated as consistent, in the center of the distribution (20 percentile to 80 percentile), and the others two, as under- and over-estimated populations, representing the tails of the distribution. For under- and over-estimated categories, the vertical structure of the clouds was evaluated. The underestimation error is correlated with almost all cloud properties (rain rate, cloud top, Liquid Water path (LWP), Ice Water Path (IWP), polarization and Polarization Corrected Temperature at 85 GHZ (PCT85)) while the overestimation error is only function of the IWP. The use of combinations of low and high frequency channels was able to identify some characteristics associated with under- and over-estimated cases. A high positive difference between the 10 GHz and 85 GHz as well as 19 GHz – 85 GHz is characteristic of very high scattering at 85 GHz (high amount of ice) and small liquid water amounts corresponding to cases that are often overestimated by the radiometer. On the other hand, underestimated cases have smaller ice particles that are not sensitive to the high frequency microwaves channels measured by TMI. These results open potential new avenues to improve the quality of passive microwave rainfall estimates.

A comparison of zonotope order reduction techniques

Abstract This brief paper provides a comparison of methods for enclosing a given zonotope within another of lower complexity, commonly called order reduction. These techniques are essential for maintaining efficiency in recursive computations with zonotopes and are widely used in set-based estimation, hybrid systems verification, and fault detection. We first review existing methods and provide a new theoretical analysis of the method recently introduced by Scott et al. (2016). We then compare methods in terms of computational cost and overestimation error, and investigate the effects of zonotope dimension, initial order, and reduced order on these metrics. These results provide valuable guidance for the design of robust estimation and control algorithms that more effectively balance accuracy with computational cost.

Trajectory analysis of land use and land cover maps to improve spatial–temporal patterns, and impact assessment on groundwater recharge

Land use/land cover (LULC) change is a consequence of human-induced global environmental change. It is also considered one of the major factors affecting groundwater recharge. Uncertainties and inconsistencies in LULC maps are one of the difficulties that LULC timeseries analysis face and which have a significant effect on hydrological impact analysis. Therefore, an accuracy assessment approach of LULC timeseries is needed for a more reliable hydrological analysis and prediction. The objective of this paper is to assess the impact of land use uncertainty and to improve the accuracy of a timeseries of CORINE (coordination of information on the environment) land cover maps by using a new approach of identifying spatial–temporal LULC change trajectories as a pre-processing tool. This ensures consistency of model input when dealing with land-use dynamics and as such improves the accuracy of land use maps and consequently groundwater recharge estimation. As a case study the impact of consistent land use changes from 1990 until 2013 on groundwater recharge for the Flanders-Brussels region is assessed. The change trajectory analysis successfully assigned a rational trajectory to 99% of all pixels. The methodology is shown to be powerful in correcting interpretation inconsistencies and overestimation errors in CORINE land cover maps. The overall kappa (cell-by-cell map comparison) improved from 0.6 to 0.8 and from 0.2 to 0.7 for forest and pasture land use classes respectively. The study shows that the inconsistencies in the land use maps introduce uncertainty in groundwater recharge estimation in a range of 10–30%. The analysis showed that during the period of 1990–2013 the LULC changes were mainly driven by urban expansion. The results show that the resolution at which the spatial analysis is performed is important; the recharge differences using original and corrected CORINE land cover maps increase considerably with increasing spatial resolution. This study indicates that improving consistency of land use map timeseries is of critical importance for assessing land use change and its environmental impact.

English

The performances of five empirical models, namely: Hargreaves-Samani, Makkink1 (1957), Makkink2 (1984), Priestley-Taylor and FAO 56-PM in estimating reference evapotranspiration (REF-ET) were separately compared with Epan data and FAO 56-PM, respectively. Based on statistical analysis, Hargreaves-Samani method compared best with daily and monthly Epan data, while Makkik2 (1984) ranked first with FAO 56-PM. In terms of regression analysis, Priestley-Taylor performed best with daily FAO 56-PM method while Hargreaves-Samani ranked first with daily Epan data. Hargreaves-Samani also correlated best with mean monthly Epan data. The quantitative evaluation of cumulative daily and monthly reference-evapotranspiration (RET-ET) values showed that Makkink (1984) produced the least overestimation and percent relative error against FAO 56-PM while Hargreaves-Samani performed best with Epan data with the least overestimation and percent relative error. In terms of cumulative monthly ETo totals for the farming season (Dec-April) over the study period, Hargreaves-Samani ranked best with Epan data with the least overestimation and percent relative error while Priestley–Taylor ranked best with FAO 56-PM producing the least overestimation. Overall, Hargreaves-Samani with its original coefficient was adjudged best, capable of approximating FAO 56-PM and Epan data in the Lower Niger River Basin, followed by Makkink (1984) and Priestley-Taylor. Penman-Monteith estimates were used to develop monthly correction factors for adjusting Empirical models for their potential use in Lower Niger Basin. A comparative study such as this has not been undertaken in the Lower Niger River Basin. The models recommended in this study are economical, lesser-data demanding and can be applied to predicting REF-ET in remote agricultural areas. Key words: Reference-evapotranspiration (RET-ET), empirical models, radiation-based methods, temperature-based methods, FAO 56 –PM, Lower Niger River Basin.

Evaluation of snow cover and snow depth on the Qinghai–Tibetan Plateau derived from passive microwave remote sensing

Abstract. Snow cover on the Qinghai–Tibetan Plateau (QTP) plays a significant role in the global climate system and is an important water resource for rivers in the high-elevation region of Asia. At present, passive microwave (PMW) remote sensing data are the only efficient way to monitor temporal and spatial variations in snow depth at large scale. However, existing snow depth products show the largest uncertainties across the QTP. In this study, MODIS fractional snow cover product, point, line and intense sampling data are synthesized to evaluate the accuracy of snow cover and snow depth derived from PMW remote sensing data and to analyze the possible causes of uncertainties. The results show that the accuracy of snow cover extents varies spatially and depends on the fraction of snow cover. Based on the assumption that grids with MODIS snow cover fraction &gt; 10 % are regarded as snow cover, the overall accuracy in snow cover is 66.7 %, overestimation error is 56.1 %, underestimation error is 21.1 %, commission error is 27.6 % and omission error is 47.4 %. The commission and overestimation errors of snow cover primarily occur in the northwest and southeast areas with low ground temperature. Omission error primarily occurs in cold desert areas with shallow snow, and underestimation error mainly occurs in glacier and lake areas. With the increase of snow cover fraction, the overestimation error decreases and the omission error increases. A comparison between snow depths measured in field experiments, measured at meteorological stations and estimated across the QTP shows that agreement between observation and retrieval improves with an increasing number of observation points in a PMW grid. The misclassification and errors between observed and retrieved snow depth are associated with the relatively coarse resolution of PMW remote sensing, ground temperature, snow characteristics and topography. To accurately understand the variation in snow depth across the QTP, new algorithms should be developed to retrieve snow depth with higher spatial resolution and should consider the variation in brightness temperatures at different frequencies emitted from ground with changing ground features.