Aging profoundly alters the neuromotor and cognitive systems that support gait control, leading to increased variability and instability that predict functional decline and dementia risk. In this pilot study, conducted to inform the design of the Semmelweis Study gait assessment pipeline, we examined how aging and cognitive load influence the magnitude and temporal organization of gait fluctuations. The Semmelweis Study is a large, prospective workplace cohort at Semmelweis University designed to identify the determinants of unhealthy aging and the mechanisms that preserve functional resilience across the life course. One hundred three adults aged 23-87years completed single- and dual-task walking trials on a 20-foot pressure-sensitive walkway. Gait variability was quantified using the median absolute deviation (MAD) and coefficient of variation (CoV) of key spatiotemporal parameters, while permutation entropy (PE) captured the complexity of stride-to-stride dynamics. Aging was associated with progressive increases in both the variability (MAD, CoV) and changes in orderliness (PE) of gait fluctuations, particularly under dual-task conditions, suggesting a dual contribution of neuromotor degradation and compensatory recruitment of higher-order control processes. The amplification of these effects during cognitive load highlights the vulnerability of cognitive-motor integration with advancing age. By integrating robust, relative, and nonlinear variability metrics within a unified analytical framework, this study provides a multidimensional characterization of gait control and establishes sensitive indicators for detecting early functional decline. Within the translational framework of the Semmelweis Study, these quantitative gait measures-together with vascular, metabolic, and cognitive assessments-are expected to serve as informative components of a comprehensive biomarker system aimed at identifying early determinants of unhealthy brain aging and guiding preventive strategies to promote healthy longevity.

Robust Classification of Stock Market Volatility Using Median Absolute Deviation: Evidence from Global Indices

Abstract Precise measurements of Type Ia supernovae (SNe Ia) at low redshifts ( z ) serve as one of the most viable keys to unlocking our understanding of cosmic expansion, isotropy, and growth of structure. The Dark Energy Bedrock All-Sky Supernovae (DEBASS) program will deliver a uniformly calibrated low- z dataset of more than 400 spectroscopically confirmed SNe Ia in the Southern Hemisphere. DEBASS utilizes the Dark Energy Camera to image supernovae in conjunction with the Wide-Field Spectrograph to gather comprehensive host-galaxy information. By using the same photometric instrument as both the Dark Energy Survey (DES) and the DECam Local Volume Exploration Survey, DEBASS not only benefits from a robust photometric pipeline and well-calibrated images across the Southern sky, but can replace the historic and external low- z samples that were used in the final DES supernova analysis. In this paper, along with a companion paper, we present an early data release of 77 DEBASS SNe within the DES footprint. We introduce the DEBASS program, discuss its scientific goals and the advantages it offers for supernova cosmology, and present our initial results demonstrating data quality. With this early data release, we find a robust median absolute standard deviation of Hubble diagram residuals of ∼0.10 mag and an initial measurement of the host-galaxy mass step of 0.06 ± 0.04 mag, both before performing bias corrections. This low scatter shows the promise of a low- z SN Ia program with a well-calibrated telescope and high signal-to-noise ratio across multiple bands.

The arid alpine landscape on the northern margin of the Qaidam Basin, northern Tibetan Plateau, presents a challenge for conventional geochemical background modeling. Leveraging a unique multi-scale geochemical dataset (1:200 000, 1195 samples; 1:50 000, 18 855 samples; 1:25 000, 35 675 samples) from the Qinghai Geological Exploration Fund, this study systematically evaluates scale effects on background estimation. We compared the performance of four prevalent methods – iterative exclusion, median absolute deviation, exploratory data analysis, and the concentration–area fractal method – and found that a multi-method approach effectively balances the comprehensiveness and precision of the background modeling framework. To address the spatial heterogeneity of backgrounds caused by complex geology, we propose a novel subzone robust background method (SRBM). This method calculates robust median background values within distinct geological units and generates an adaptive background field through area-weighted fusion. In the Banhongshan area, application of the SRBM precisely calibrated the gold background to 0.47 ng g −1 , successfully identifying two concealed gold anomalies obscured by traditional methods; one anomaly coincides perfectly with known industrial orebodies. This research demonstrates that high-density sampling (1:25 000) significantly enhances the signal-to-noise ratio for chalcophile elements in arid alpine terrains. The SRBM, by integrating geological knowledge with robust statistics, provides a powerful and reliable tool for weak geochemical signal extraction in both mineral exploration and environmental assessment.

The transverse beam size is a key parameter for characterizing the performance of synchrotron radiation sources. Accurate measurement of the transverse beam size is crucial for assessing beam quality. In this study, a fiber array-photomultiplier tube (PMT) beam measurement system was developed to enable high-precision sampling of beam profile information for beam-size measurement. Furthermore, a hybrid method integrating nonlinear least squares (NLLS) fitting and Gaussian basis-function fitting was proposed to reconstruct the beam intensity profile from discrete sampling data. Before performing NLLS fitting, a median absolute deviation (MAD)-based threshold filter is employed to remove outliers and suppress random noise, thereby improving the stability and robustness of the parameter estimation. The filtered data are then fitted using NLLS to obtain the reconstructed distribution. To capture potential high-order modal features in the beam profile, a Gaussian basis-function fitting model was also introduced for comparison, and its performance was evaluated under complex intensity distributions. Additionally, the relationship between the full width at half maximum (FWHM) and beam intensity was experimentally verified while accounting for measurement effects in the system. The results demonstrate that the proposed hybrid algorithm improves reconstruction accuracy and robustness, enabling precise recovery of the beam-intensity profile in the fiber-array PMT system.

With the increased challenges in diagnosing DDH using traditional ultrasound imaging methods, accurate diagnosis is essential. This study assesses the effectiveness of AI in the imaging-based diagnosis of DDH through a systematic review. This review was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and registered in Prospero (registration ID: CRD42024563606). A comprehensive search was conducted across Ovid MEDLINE, PubMed, Embase, the Cochrane Central Register of Controlled Trials, and the Cochrane Database of Systematic Reviews. Studies were screened using selection criteria, and quality was assessed using standardised tools. Thematic content analysis was also performed. Of the 32 studies identified, 19 were included, with 15 undergoing quantitative analysis. The main outcome measures were sensitivity, specificity, accuracy, AUROC, PPV, and NPV. Median, median absolute deviation, Bonett-Price 95% confidence intervals, maximum, minimum, and interquartile ranges were calculated and presented in a box-and-whisker diagram. In the 19 included studies, the median sensitivity was 90.0% and specificity was 93.2% across 36,907 patients. Fifteen studies reported diagnostic accuracy, with a median of 92.6%. Accuracy rates ranged from 79.2 to 99%. The most common model architecture was mask R-CNN. Four studies (21%) were judged to have a high risk of bias using the QUADAS-2 tool. AI technologies hold significant potential for enhancing the diagnostic accuracy of DDH. However, existing variability and bias across studies highlight the need for further standardisation and validation.

A damage localisation method for bridge structures based on the fusion of multi-source vehicle-induced responses is proposed. The proposed method addresses the challenges of insufficient utilisation of dynamic response information and the large amount of labelled data required by supervised learning models, aiming to improve the accuracy of damage localisation. Firstly, acceleration and displacement signals are obtained from multiple sensor locations as a moving vehicle passes over a bridge and fed into a deep convolutional autoencoder (DCAE), which automatically extracts damage-sensitive features without requiring labelled data. Furthermore, the damage index is calculated based on the difference between bridge damage-sensitive features before and after damage, using the median absolute deviation (MAD) standardisation process. Finally, Dempster-Shafer (D-S) evidence theory is applied to integrate multi-source data, and the damage probability at each sensor is calculated for damage localisation. The finite element model is employed for numerical validation to simulate damage scenarios under various noise levels and damage severities, and effectiveness of the proposed method is validated through laboratory tests on a scaled steel-concrete composite bridge subjected to a two-axle vehicle. The results demonstrate that the proposed method accurately identifies damage location(s) for both single and multiple damage scenarios utilising the vehicle-induced responses.

Abstract Solid pseudopapillary neoplasm (SPN) of the pancreas is a rare, low-grade malignant tumor that predominantly affects young women. Although previous studies have applied omics approaches, including whole-exome sequencing, transcriptomics, and DNA methylation profiling to characterize SPN, its tumorigenesis and cell of origin remain elusive. Apart from the recurrent CTNNB1 hotspot mutation, which is currently considered the only canonical driver, SPN is typically diploid and exhibits few recurrent genomic alterations. Moreover, because of its low incidence, previous small-cohort studies have been insufficient for robust molecular subtyping or integrative analysis. To comprehensively characterize the molecular landscape of SPN, we performed multi-omics analyses on 80 tumors using whole-genome sequencing (WGS), whole-transcriptome sequencing, and whole-genome enzymatic methyl-sequencing (EM-seq). Unsupervised hierarchical clustering based on the top 1,000 genes ranked by median absolute deviation (MAD) across transcriptomes identified two robust molecular subtypes: a developmentally reprogrammed type and an immune-enriched type: The immune-enriched subtype exhibited significant activation of immune-associated pathways, including immunoglobulin-mediated and adaptive immune responses, whereas the developmentally reprogrammed subtype showed enrichment of embryonic skeletal system development and morphogenesis signatures. Furthermore, SPN tumors displayed a genome-wide hypermethylated pattern compared with normal pancreatic tissues, highlighting their distinct epigenetic state. Beyond expression-based stratification, we systematically examined fusion events, genome-wide somatic mutations, structural variants, mitochondrial genome alterations, numtogenesis, retrotransposon activity, and methylation entropy across the cohort. Notably, a recurrent TVP23C-CDRT4 fusion was detected in 61% of cases, suggesting a potential novel genomic hallmark of SPN. This study represents the largest integrative multi-omics analysis of SPN to date, revealing distinct transcriptomic and epigenomic subtypes, a recurrent fusion event, and marked molecular heterogeneity. Ongoing integrative analyses aim to clarify the developmental origin and oncogenic evolution of this rare pancreatic neoplasm. Citation Format: Dong-Ju Shin, Jin Ho Choi, Sang Hyub Lee, In Rae Cho, Kyung-Min Lee, Ji Kon Ryu, Woo Hyun Paik, Jin-Ku Lee. Comprehensive multi-omics profiling reveals molecular heterogeneity and developmental signatures in solid pseudopapillary neoplasm of the pancreas [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7261.

The lack of a robust long-term environmental baseline has hindered risk management in the industrial complex region of Abaetetuba and Barcarena, Eastern Amazon. This study fills that gap by analyzing a 41-year time series (1980-2021; N = 19,292) and applying statistical tools such as the Median Absolute Deviation (MAD) to establish the geochemical background. Results for key physicochemical parameters indicate typical tropical conditions (temperature near 30°C; slightly acidic pH: 6.8 ± 1.0) and low buffering capacity (alkalinity < 30mg l⁻1). Nutrient data revealed clear anthropogenic enrichment, with total phosphorus (TP) occurring in excess (0.2 ± 1.9mg l⁻1) and a low N/P ratio (2.5). Four metallic elements exceeded legal thresholds: aluminum (Al, 0.0009-1.31mg l⁻1), iron (Fe, 0.002-1.14mg l⁻1), cadmium (Cd, 0.0001-0.001mg l⁻1), and lead (Pb, 0.0001-0.019mg l⁻1). While Al and Fe concentrations are linked to regional geology and exhibited negative temporal trends, Cd and Pb require particular attention due to their association with rising anthropogenic pressures. Overall, the study successfully identified environmental thresholds, generating the first 41-year background reference for the region. This contribution is crucial for distinguishing natural variability from long-term contamination and significantly strengthens environmental monitoring and impact assessment in Amazonian aquatic ecosystems.

Label-free, single-cell-precise, and monomeric-unit-resolved monitoring of biopolymer fermentation by ramanomics.

There is limited literature on the ability of high-frequency cerebral functional near-infrared spectroscopy (fNIRS) systems to characterize cerebral autoregulation/cerebrovascular reactivity (CA/CVR) regional disparity than other low-frequency commercial systems. To overcome temporal and spatial limitations of existing commercial NIRS systems, we created a custom-built whole brain CVR mapping system using fNIRS. We preliminarily evaluated regional hemispheric disparity in CA/CVR using various fNIRS derived metrics based on relative hemoglobin concentrations. Healthy volunteer data was recorded for approximately 90 minutes in a block-trial fashion with baseline and perturbation testing. Five types of hemoglobin-based indices were derived using 1 Hz and 250 Hz sampled data. Regional hemispheric disparity between brain lobar regions was evaluated based on median and median absolute deviation metrics. Multi-variate cerebral physiologic relationships between hemispheres were assessed via optimal autoregressive integrative moving average (ARIMA), vector ARIMA impulse response functions, and Granger causality analyses. Data from 50 healthy control volunteers were prospectively analyzed. Our system was able to detect subtle differences between corresponding right and left brain regions through all statistical methods employed, demonstrating the ability of our novel system to detect changes in regional variation of fNIRS and derived CVR measures. These were present and magnified during perturbation testing compared to baseline recordings. However, given the healthy nature of the study population, large differences in CVR measures between brain regions and extreme CVR derangements were not seen. Our custom built fNIRS whole brain CVR mapping system was able to detect subtle regional differences in CVR measures across various time-domain analytic techniques. These findings are in alignment with prior literature, supporting the notion that research-grade fNIRS systems may be adequate for regional disparity analysis of CVR in humans. Future work in diseased/injured human cohorts is required to further quantify the sensitivity of our custom-built system to detect regional variations and disturbances in CVR.

Stable Diffusion is a widely used text-to-image generation model. However, its outputs are highly sensitive to hyperparameters settings and often suffer issues such as semantic drift, subject misalignment and detail loss. Traditional methods rely on manually adjusting hyperparameters to alter the attention distribution and thus improve the quality of generated images, which is time-consuming and lacks precision. Therefore, we propose an attention-guided visual diagnostic system named DAttnVis, which is designed to assist users in understanding the complex inference process of the Stable Diffusion model and optimizing its parameters. The core idea is to transform high-dimensional attention signals into comparable diagnostic representations across layers using a quantifiable metric—the Attention Concentration Index (ACI). Additionally, an anomaly detection method based on Median Absolute Deviation (MAD) is proposed to accurately identify abnormal attention layers. By linking multiple views, including UNet attention flow, diagnosis and guidance, cross-attention, and historical comparison, DAttnVis constructs a comprehensive diagnostic workflow that covers global screening, structural drilling-down, semantic tracing, and result verification. Quantitative evaluation experiments, case studies and user studies demonstrate that DAttnVis can effectively reduce trial-and-error costs and debugging burdens in the model tuning process, while improving the accuracy of anomalous structure localization and key prompt attribution.

High-spatiotemporal-resolution land surface temperature (LST) data are crucial for analyzing surface energy balance, modeling temperature-related processes, and monitoring thermal environments. However, despite advancements in multi-source fusion and reconstruction techniques, high-frequency LST data remain susceptible to anomalies such as abrupt changes and outliers due to retrieval uncertainties and varying observation conditions. Conventional statistical outlier detection methods risk misidentifying physically plausible rapid weather changes as data errors, introducing systematic biases. To address this, we propose a two-stage anomaly detection framework that follows a “temporal physical pre-screening first, spatial statistical verification later” logic. First, a piecewise empirical model, based on typical diurnal LST variation characteristics, is constructed to identify points violating physical patterns. Subsequently, a spatial consistency test using median absolute deviation (MAD) is introduced to distinguish real weather-driven fluctuations from genuine data anomalies from a spatial synergy perspective. This sequential design effectively reduces the risk of mis-correcting physically reasonable temperature variations. Validated using hourly seamless LST data (2016–2021) and ground observations in the Heihe River Basin, our method outperformed Seasonal-Trend decomposition using Loess (STL), double standardization methods, and robust Holt–Winters. For over 87% of the detected anomalies, the proposed method demonstrated positive improvement rates in RMSE, MAE, R, and R2. The overall average improvement rates reached 23.61%, 18.79%, 16.46%, and 61.33%, respectively, indicating robust performance. The results underscore that explicitly incorporating physical constraints enhances the reliability and interpretability of quality control for high-temporal-resolution remote sensing LST data.

Abstract To address the challenge of estimating redshifts when only single-band images are available, this study introduces a deep learning model named Vision Transformer (ViT)-MDNz. Leveraging robust statistical priors learned from large-scale data concerning the correlation between redshift and morphology, the model can directly estimate redshifts and their associated uncertainties from single-band galaxy images. It integrates a ViT to extract deep morphological features and a mixture density network (MDN) to predict the full redshift probability density function (PDF). Trained and evaluated on approximately 300,000 single-band images from the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys (DESI-LS), the model achieves a normalized median absolute deviation σ NMAD = 0.034 and an outlier fraction f out = 2.6% in the r band for redshifts up to z ≲ 1. Evaluations using probability integral transform and continuous ranked probability score confirm that the predicted PDFs are well calibrated and closely match the true distribution. These results demonstrate that competitive redshift estimates can be obtained using morphological features alone, and that incorporating color information further enhances the accuracy and robustness of the estimation. Therefore, ViT-MDNz provides a practical approach for redshift estimation of galaxy samples with limited photometric band coverage, contributing to improved completeness and usability of redshift catalogs for future large-scale surveys such as DESI and Legacy Survey of Space and Time.

Recent spectroscopic surveys provide element abundances for large samples of Milky Way stars, from which stellar parameters can be inferred. Stellar ages, among them, are both a notoriously difficult parameter to estimate and a fundamental property for Galactic archaeology studies. We aim to develop a model-driven deep learning approach to age determination by training neural networks on stellar evolutionary grids. Contrary to the usual data-driven deep learning approach of using prior age estimates as training data, our method has the potential for a wider and less biased range of application. The low computational cost of deep learning methods compared to, for example, Bayesian isochrone fitting enables a broad analysis of large spectroscopic catalogues. We trained multilayer perceptrons on different stellar evolutionary grids to map M/H , M_G, $(G_ BP -G_ RP )$ to stellar age τ. We combined Gaia photometry and parallaxes, metallicities, and α elements from spectroscopic surveys and extinction maps, which are passed through neural networks to estimate stellar ages. We applied our method to the LAMOST DR10, GALAH DR3 &amp; DR4, and APOGEE DR17 spectroscopic surveys, estimating ages using the BaSTI tracks and other stellar evolutionary models. We leveraged this novel technique to study, for the first time, differences in age estimates from several evolutionary grids applied to very large datasets. In addition, we dated 13 open clusters and one globular cluster, finding a median absolute deviation with literature ages of 0.20 Gyr. Along with the stellar age catalogues from our estimates, we release (Neural Estimator of Stellar Times), a python package to estimate stellar age based on this work, as well as a web interface. NEST We show that, when using the same evolutionary grid, our method retrieves the same ages as a Bayesian approach similar to SPInS, for only a fraction of the computational cost, with a 60,000 speed-up factor for a typical star. This model-driven deep learning technique thus opens up the way for broad galactic archaeology studies on the largest datasets available today and in the near future with upcoming surveys such as 4MOST.

ABSTRACT We present a new set of Rayleigh-wave detection algorithms (a module) that we derive from two competing, approximate models for elliptically polarized seismic data. This module processes three-component seismograms with sliding windows to output estimates of test statistics and source back azimuths that it can combine from multiple frequency bands. The module automatically adjusts declaration thresholds to maintain a fixed false alarm rate against noise and adapts its sliding window lengths to include a fixed number of waveform cycles per frequency band. We demonstrate these capabilities against real Rayleigh waves that are sourced from airborne explosions in Ukraine. The module shows reliable detection rates against explosions and reasonable estimates of source back azimuths at computational costs akin to those of power detectors. Performance curves show that the detection module exceeds a 0.80 true positive detection rate against a real, ∼75 kg Trinitrotoluene yield equivalent source at ranges of 25 km. Back-azimuthal estimates achieve mean errors of ∼3° with median absolute deviations of ∼10°. Our module thereby demonstrates a capability to automatically detect and directionally locate airborne explosions from three-component seismic data, simultaneously. Plain Language Summary: Rayleigh waves are a type of seismic surface wave that moves each point on the ground in the shape of an ellipse. This article develops and tests a new set of algorithms, or a module, to detect such Rayleigh waves with sensors that record ground motion in the east, north, and vertical directions. Our module operates like the scan feature of traditional car radios, but against seismic data. During its “scan” operation, our module consistently tests a fixed number of waveform cycles for Rayleigh-wave motion and adjusts its sensitivity to falsely trigger fewer than a set number of times per year. Our module also estimates the direction from which the source of this motion came. We test our module against data that we predict from airborne explosion simulations and from real explosions recorded in Ukraine. We thereby detect real events with the energy of about 75 kg of Trinitrotoluene explosive from 25 km away, about 80 out of 100 times. We also correctly estimate the direction back to the source within about 10°. Our article and its supplemental material provide the mathematics, physics, and some data required to understand and use this module.

Clonal expansion (CE) of cells carrying cancer driver mutations (CDMs) is being developed as a biomarker of cancer risk. CE in lung of MutaMouse males treated with 0, 6.25, 12.5, and 25 mg/kg/d benzo[b]fluoranthene (B[b]F) by gavage for 90 and 180 d was assessed by CarcSeq. DNA regions encompassing mouse correlates of human hotspot CDMs were PCR amplified, attaching 18-base unique molecular identifiers (UMIs) during the PCR. Following library preparation and sequencing, UMI-defined read families were assembled to produce single-strand consensus sequences (SSCSs). Recovered mutants with mutant fractions (MFs) ≥10−4 were stratified based on their occurrence in lung-specific or nonlung driver sequences and CE was assessed on a per mouse basis as median absolute deviation in mutant fraction (MAD). A significant, dose-dependent increase in MAD was observed for lung-specific MFs after 180 d of B[b]F treatment, a duration that did not cause a significant increase in lung lesions. Dose- and treatment duration-related increases in MF were observed for Egfr, the mouse correlate of a known human lung tumor driver gene. MF and mutation counts were significantly decreased in response to longer treatment duration for some nonlung drivers, suggesting negative selection. Importantly, the normalized trinucleotide mutation spectrum derived from CDMs reflects amplification of preexisting spontaneous mutations, distinct from those induced by B[b]F mutagenesis. These results show CarcSeq detects CE of preexisting cancer driver gene mutants induced by the genotoxic carcinogen B[b]F and suggest a CE endpoint may be useful for evaluating cancer risk associated with tumor promoters or complete carcinogens.

Outlier detection is an important prerequisite for achieving accurate Precise Point Positioning (PPP) solutions. Without a suitable procedure for identifying outliers, unexpected effects such as divergence or degradation may arise during parameter estimation. To ensure reliable positioning, this study proposes a new robust quality control strategy based on the median absolute deviation (MAD) for detecting and handling outliers, employing predicted residuals (innovations) as input. In the proposed method, innovation vectors of code and phase observations are evaluated separately. Following the detection and identification steps, outlier code observations are removed from the filter update, whereas phase outliers are handled by reinitializing their ambiguity parameters through variance inflation in the predicted covariance matrix. Moreover, an existing quality control algorithm incorporating both posterior and standardized residuals was adapted and tested as the second method. The robust adaptive Kalman filter with the IGG (Institute of Geodesy and Geophysics) III function was then applied for a performance comparison with the two proposed methods. In this regard, GNSS observations collected during static and kinematic experimental tests conducted with a low-cost GNSS receiver were processed using offline real-time products of the Multi-GNSS Advanced Demonstration tool for Orbit and Clock Analysis (MADOCA). The positioning accuracies for the static tests demonstrated that the IGG-III robust adaptive Kalman filter is inferior to the presented strategies and exhibits substantial solution deviations, especially when there is limited visibility of the satellites. While multi-GNSS combinations significantly improved the performance of this method, they did not reach the sub-decimeter 3D RMS delivered by our proposed techniques. Additionally, the kinematic experiment showed that the two methods produced comparable results and were superior to the IGG-III. Indeed, these approaches achieved a 3D accuracy of about 12 cm, whereas the IGG-III method yielded an accuracy of 17.7 cm. As a result of static open-sky environment tests and a kinematic suburban experiment, the MAD-based method, which is straightforward to implement, achieved strong performance in detecting outliers and contributed to the promising potential of low-cost receivers. The findings demonstrate that the proposed method can support practical applications, such as vehicle navigation, precision agriculture, and precise positioning of UAVs/drones.

• Integrating long-term seasonal data clarifies climate change impacts • The strongest signs of climate change are observed in autumn • Rising summer/autumn LST spans 4,315 km² and 2,762 km² of suitable habitat • Vegetation health anomalies cover more suitable habitat than temperature anomalies • Topographic parameters govern microclimates in population cores across seasons Environmental pressures demand effective habitat conservation through systematic, long-term monitoring of key ecological indicators. In this study, seasonal averages of Land Surface Temperature (LST) and the Normalized Difference Vegetation Index (NDVI) were derived from MODIS data (2003–2024) and categorized into winter, spring, summer, and autumn. LST data were downscaled using Geographically Weighted Regression (GWR) with NDVI, while the Vegetation Health Index (VHI)—a drought indicator—was analyzed. Long-term seasonal trends and seasonal anomalies were identified using the Mann-Kendall test (95% significance) and median absolute deviation, with the Elbow method quantifying anomaly extents. Seasonal cycles were further evaluated by comparing standardized Z-score values, and long-term microclimate patterns were characterized through Principal Component Analysis (PCA) of time-series VHI, LST, and topographic variables. Additionally, Random Forest Regression (RFR) assessed the influence of environmental factors on microclimate fluctuations. Findings reveal that increasing summer and autumn LST trends affected 4,314.78 and 2,761.12 km² of suitable habitat, with 58 and 33 population cores exhibiting rising trends, respectively. Fall is warming up, and southern demographics are experiencing rising trends in LST and VHI. Notably, LST and VHI anomalies covered smaller habitat areas than overall LST trends, with the sharpest long-term seasonal shifts observed from winter to autumn. RFR results indicate that these microclimates are primarily driven by topography-related factors, including the Compound Topographic Index and openness measures. Our integrated analysis of vegetation and temperature trends offers a robust framework for conservation planning by pinpointing habitats most vulnerable to climate-induced changes.

Abstract The Bonneville Salt Flats (BSF) in northwestern Utah, USA, has experienced changes in area, thickness, and hydrology over the past century. This study investigates the relationship between multispectral Halite Index (HI) values, which are sensitive to halite moisture content, from Landsat and Sentinel‐2 imagery and groundwater depth (GWD) measured by nine piezometers, based on the principle that surface moisture in the halite crust is driven primarily by depth of groundwater and secondarily by atmospheric precipitation or humidity. Linear regressions reveal moderately strong correlations, and regression coefficients are used to calibrate HI values to GWD. Satellite‐calibrated GWD time series show temporal alignment with piezometer records, with a median absolute deviation less than 15 cm for seven of the nine piezometers. Spatial interpolations of regression coefficients are used to calibrate HI imagery to GWD imagery, highlighting the groundwater table's spatiotemporal variability and relation to climate. Groundwater depth is observed to be very sensitive to climatic conditions, as GWD across the crust scales with Palmer Drought Severity Index data. The methodology is tested for available piezometer data at the Badwater Basin, Salar de Uyuni, and Salar de Atacama to explore global applicability. The results from this study confirm the potential of multispectral remote sensing for monitoring GWD at the BSF and suggest the methodology could be transferable to similar salt crusts globally.