Search

Climate change reduces snowpack, advances snowmelt phenology, drives summer warming, alters growing season precipitation regimes, and consequently modifies vegetation phenology in mountain systems. Elevational migrants track spatial variation in seasonal plant growth by moving between ranges at different elevations during spring, so climate-driven vegetation change may disrupt historic benefits of migration. Elevational migrants can furthermore cope with short-term environmental variability by undertaking brief vertical movements to refugia when sudden adverse conditions arise. We uncover drivers of fine-scale vertical movement variation during upland migration in an endangered alpine specialist, Sierra Nevada bighorn sheep (Ovis canadensis sierrae) using a 20-year study of GPS collar data collected from 311 unique individuals. We used integrated step-selection analysis to determine factors that promote vertical movements and drive selection of destinations following vertical movements. Our results reveal that relatively high temperatures consistently drive uphill movements, while precipitation likely drives downhill movements. Furthermore, bighorn select destinations at their peak annual biomass and maximal time since snowmelt. These results indicate that although Sierra Nevada bighorn sheep seek out foraging opportunities related to landscape phenology, they compensate for short-term environmental stressors by undertaking brief up- and downslope vertical movements. Migrants may therefore be impacted by future warming and increased storm frequency or intensity, with shifts in annual migration timing, and fine-scale vertical movement responses to environmental variability.

Increasing wildfire and declining snowpacks in mountain regions threaten water availability. We combine satellite-based fire detections with snow seasonality classifications to examine fire activity in California's seasonal and ephemeral snow zones. We find a nearly tenfold increase in fire activity during 2020–2021 versus 2001–2019. Accumulation season broadband snow albedo declined 25%–71% at two burned sites (2021 and 2022) according to in-situ data relative to un-burned conditions, with greater declines associated with increased burn severity. By enhancing snowpack susceptibility to melt, both decreased snow albedo and canopy drove midwinter melt during a multi-week dry spell in 2022. Despite similar meteorological conditions in December–February 2013 and 2022–linked to persistent high pressure weather regimes–minimal melt occurred in 2013. Post-fire snowpack differences are confirmed with satellite measurements. With growing geographical overlap between wildfire and snow, our findings suggest California's snowpack is increasingly vulnerable to the compounding effects of dry spells and wildfire.

Abstract. Given the tradeoffs between spatial and temporal resolution, questions aboutresolution optimality are fundamental to the study of global snow. Answersto these questions will inform future scientific priorities and missionspecifications. Heterogeneity of mountain snowpacks drives a need for dailysnow cover mapping at the slope scale (≤30 m) that is unmet for avariety of scientific users, ranging from hydrologists to the military towildlife biologists. But finer spatial resolution usually requires coarsertemporal or spectral resolution. Thus, no single sensor can meet all theseneeds. Recently, constellations of satellites and fusion techniques havemade noteworthy progress. The efficacy of two such recent advances isexamined: (1) a fused MODIS–Landsat product with daily 30 m spatialresolution and (2) a harmonized Landsat 8 and Sentinel 2A and B (HLS) product with3–4 d temporal and 30 m spatial resolution. State-of-the-art spectral unmixingtechniques are applied to surface reflectance products from 1 and 2 tocreate snow cover and albedo maps. Then an energy balance model was run toreconstruct snow water equivalent (SWE). For validation, lidar-basedAirborne Snow Observatory SWE estimates were used. Results show thatreconstructed SWE forced with 30 m resolution snow cover has lower bias, ameasure of basin-wide accuracy, than the baseline case using MODIS (463 mcell size) but greater mean absolute error, a measure of per-pixelaccuracy. However, the differences in errors may be within uncertaintiesfrom scaling artifacts, e.g., basin boundary delineation. Other explanationsare (1) the importance of daily acquisitions and (2) the limitations ofdownscaled forcings for reconstruction. Conclusions are as follows: (1) spectrallyunmixed snow cover and snow albedo from MODIS continue to provide accurateforcings for snow models and (2) finer spatial and temporal resolutionthrough sensor design, fusion techniques, and satellite constellations arethe future for Earth observations, but existing moderate-resolution sensorsstill offer value.

Whereas many independent methods are used to estimate snow water equivalent (SWE) and its spatial distribution and seasonal variability, a need exists for a systematic characterization of inter-model differences at annual, seasonal, and regional scales necessary to quantify the associated uncertainty in these datasets. This study conducts a multi-scale validation and comparison, based on Airborne Snow Observatory data, of five state-of-the-art SWE datasets in the Sierra Nevada, California, including three SWE datasets from retrospective models: an INiTial REConstruction model (REC-INT), an improved REConstruction model based on the ParBal energy balance model (REC-ParBal), and a Sierra Nevada SWE REConstruction with Data Assimilation (REC-DA), and two operational SWE datasets from the U.S. National Weather Service, including the Snow Data Assimilation System (SNODAS) and the National Water Model (NWM-SWE). The results show that REC-DA and REC-ParBal provide the two most accurate estimates of SWE in the snowmelt season, both with small positive biases. REC-DA provides the most accurate spatial distribution of SWE (R2 = 0.87, MAE = 66 mm, PBIAS = 8.3%) at the pixel scale, while REC-ParBal has the least basin-wide PBIAS (R2 = 0.79, MAE = 73 mm, PBIAS = 4.1%) in the snowmelt season. Moreover, REC-DA underestimates peak SWE by −5.8%, while REC-ParBal overestimates it by 7.5%, when compared with the measured peak SWE at snow pillow stations across the Sierra Nevada. The two operational SWE products—SNODAS and NWM-SWE—are less accurate. Furthermore, the inter-model comparison reveals a certain amount of disagreement in snow water storage across time and space between SWE datasets. This study advances our understanding of regional SWE uncertainties and provides critical insights to support future applications of these SWE data products and therefore has broad implications for water resources management and hydrological process studies.

Abstract. Given the tradeoffs between spatial and temporal resolution, questions about resolution optimality are fundamental to the study of global snow. Answers to these questions will inform future scientific priorities and mission specifications. Heterogeneity of mountain snowpacks drives a need for daily snow cover mapping at the slope scale (≤ 30 m) that is unmet for a variety of scientific users, ranging from hydrologists to the military to wildlife biologists. But finer spatial resolution usually requires coarser temporal or spectral resolution. Thus, no single sensor can meet all these needs. Recently, constellations of satellites and fusion techniques have made noteworthy progress. The efficacy of two such recent advances is examined: 1) a fused MODIS - Landsat product with daily 30 m spatial resolution; and 2) a harmonized Landsat 8 - Sentinel 2A/B (HLS) product with 2–3 day temporal and 30 m spatial resolution. State-of-art spectral unmixing techniques are applied to surface reflectance products from 1 & 2 to create snow cover and albedo maps. Then an energy balance model was run to reconstruct snow water equivalent (SWE). For validation, lidar-based Airborne Snow Observatory SWE estimates were used. Results show that reconstructed SWE forced with 30 m resolution snow cover has lower bias, a measure of basin-wide accuracy, than the baseline case using MODIS (463 m cell size), but higher mean absolute error, a measure of per-pixel accuracy. However, the differences in errors may be within uncertainties from scaling artifacts e.g., basin boundary delineation. Other explanations are 1) the importance of daily acquisitions and 2) the limitations of downscaled forcings for reconstruction. Conclusions are: 1) spectrally unmixed snow cover and snow albedo from MODIS continue to provide accurate forcings for snow models; and 2) finer spatial and temporal resolution through sensor design, fusion techniques, and satellite constellations are the future for Earth observations.

We present the first application of the Snow Covered Area and Grain size model (SCAG) to the Visible Infrared imaging Radiometer Suite (VIIRS) and assess these retrievals with finer‐resolution fractional snow cover maps from Landsat 8 Operational Land Imager (OLI). Because Landsat 8 OLI avoids saturation issues common to Landsat 1–7 in the visible wavelengths, we re-assess the accuracy of the SCAG fractional snow cover maps from Moderate Resolution Imaging Spectroradiometer (MODIS) that were previously evaluated using data from earlier Landsat sensors. Use of the fractional snow cover maps from Landsat 8 OLI shows a negative bias of −0.5% for MODSCAG and −1.3% for VIIRSCAG, whereas previous MODSCAG evaluations found a bias of −7.6% in the Himalaya. We find similar root mean squared error (RMSE) values of 0.133 and 0.125 for MODIS and VIIRS, respectively. The Recall statistic (probability of detection) for cells with more than 15% snow cover in this challenging steep topography was found to be 0.90 for both MODSCAG and VIIRSCAG, significantly higher than previous evaluations based on Landsat 5 Thematic Mapper (TM) and 7 Enhanced Thematic Mapper Plus (ETM+). In addition, daily retrievals from MODIS and VIIRS are consistent across gradients of elevation, slope, and aspect. Different native resolutions of the gridded products at 1 km and 500 m for VIIRS and MODIS, respectively, result in snow cover maps showing a slightly different distribution of values with VIIRS having more mixed pixels and MODIS having 7% more pure snow pixels. Despite the resolution differences, the snow maps from both sensors produce similar total snow-covered areas and snow-line elevations in this region, with R2 values of 0.98 and 0.88, respectively. We find that the SCAG algorithm performs consistently across various spatial resolutions and that fractional snow cover maps from the VIIRS instruments aboard Suomi NPP, JPPS–1, and JPPS–2 can be a suitable replacement as MODIS sensors reach their ends of life.

Abstract. Intrinsic albedo is the bihemispherical reflectance independent of effectsof topography or surface roughness. Conversely, the apparent albedo is thereflected radiation divided by the incident and may be affected bytopography or roughness. For snow, the surface is often rough, and these twooptical quantities have different uses: intrinsic albedo is used inscattering equations whereas apparent albedo should be used in energybalance models. Complementing numerous studies devoted to surface roughnessand its effect on snow reflectance, this work analyzes a time series ofintrinsic and apparent snow albedos over a season at a sub-alpine site usingan automated terrestrial laser scanner to map the snow surface topography.An updated albedo model accounts for shade, and in situ albedo measurementsfrom a field spectrometer are compared to those from a spacebornemultispectral sensor. A spectral unmixing approach using a shade endmember(to address the common problem of unknown surface topography) produces grainsize and impurity solutions; the modeled shade fraction is compared to theintrinsic and apparent albedo difference. As expected and consistent withother studies, the results show that intrinsic albedo is consistentlygreater than apparent albedo. Both albedos decrease rapidly as ablationhollows form during melt, combining effects of impurities on the surface andincreasing roughness. Intrinsic broadband albedos average 0.056 greater thanapparent albedos, with the difference being 0.052 in the near infrared or0.022 if the average (planar) topography is known and corrected. Fieldmeasurements of spectral surface reflectance confirm that multispectralsensors see the apparent albedo but lack the spectral resolution todistinguish between darkening from ablation hollows versus lowconcentrations of impurities. In contrast, measurements from the fieldspectrometer have sufficient resolution to discern darkening from the twosources. Based on these results, conclusions are as follows: (1) impurity estimatesfrom multispectral sensors are only reliable for relatively dirty snow withhigh snow fraction; (2) a shade endmember must be used in spectral mixturemodels, even for in situ spectroscopic measurements; and (3) snow albedomodels should produce apparent albedos by accounting for the shade fraction.The conclusion re-iterates that albedo is the most practical snowreflectance quantity for remote sensing.

We present the first application of the Snow Covered Area and Grain size model (SCAG) to the Visible Infrared imaging Radiometer Suite (VIIRS) and assess these retrievals with finer‐resolution fractional snow cover maps from Landsat 8 Operational Land Imager (OLI). Because Landsat 8 OLI avoids saturation issues common to Landsat 1–7 in the visible wavelengths, we re-assess the accuracy of the SCAG fractional snow cover maps from Moderate Resolution Imaging Spectroradiometer (MODIS) that were previously evaluated using data from earlier Landsat sensors. Use of the fractional snow cover maps from Landsat 8 OLI shows a negative bias of −0.5% for MODSCAG and −1.3% for VIIRSCAG, whereas previous MODSCAG evaluations found a bias of −7.6% in the Himalaya. We find similar root mean squared error (RMSE) values of 0.133 and 0.125 for MODIS and VIIRS, respectively. The Recall statistic (probability of detection) for cells with more than 15% snow cover in this challenging steep topography was found to be 0.90 for both MODSCAG and VIIRSCAG, significantly higher than previous evaluations based on Landsat 5 Thematic Mapper (TM) and 7 Enhanced Thematic Mapper Plus (ETM+). In addition, daily retrievals from MODIS and VIIRS are consistent across gradients of elevation, slope, and aspect. Different native resolutions of the gridded products at 1 km and 500 m for VIIRS and MODIS, respectively, result in snow cover maps showing a slightly different distribution of values with VIIRS having more mixed pixels and MODIS having 7% more pure snow pixels. Despite the resolution differences, the snow maps from both sensors produce similar total snow-covered areas and snow-line elevations in this region, with R2 values of 0.98 and 0.88, respectively. We find that the SCAG algorithm performs consistently across various spatial resolutions and that fractional snow cover maps from the VIIRS instruments aboard Suomi NPP, JPPS–1, and JPPS–2 can be a suitable replacement as MODIS sensors reach their ends of life.

Snowpack in mountainous areas often provides water storage for summer and fall, especially in the Western United States. In situ observations of snow properties in mountainous terrain are limited by cost and effort, impacting both temporal and spatial sampling, while remote sensing estimates provide more complete spacetime coverage. Spatial estimates of fractional snow covered area (fSCA) at 30m are available every 16 days from the series of multispectral scanning instruments on Landsat platforms. Daily estimates at 463m spatial resolution are also available from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on the Terra satellite. Fusing Landsat and MODIS fSCA images creates high resolution daily spatial estimates of fSCA that are needed for various uses: to support scientists and managers interested in energy and water budgets for water resources and to understand the movement of animals in a changing climate. Here, we propose a new machine learning approach conditioned on MODIS fSCA, as well as a set of physiographic features, and fit to Landsat fSCA over a portion of the Sierra Nevada USA. The predictions are daily 30m fSCA. The approach relies on two stages of spatially‐varying models. The first classifies fSCA into three categories and the second yields estimates within (0, 100) percent fSCA. Separate models are applied and fitted within sub‐regions of the study domain. Compared with a recently‐published machine learning model (Rittger, Krock, et al., 2021), this approach uses spatially local (rather than global) random forests, and improves the classification error of fSCA by 16%, and fractionally‐covered pixel estimates by 18%. Key Points: Fusion of multiple remote sensing estimates is a promising approach to generate high resolution daily estimates of snow coverA new machine learning approach outperforms a state‐of‐the‐art competitorSpatially‐varying fitted random forests allow for geographically‐distributed feature importance [ABSTRACT FROM AUTHOR]

In the mountains, snowmelt often provides most of the runoff. Operational estimates use imagery from optical and passive microwave sensors, but each has its limitations. An accurate approach, which we validate in Afghanistan and the Sierra Nevada USA, reconstructs spatially distributed snow water equivalent (SWE) by calculating snowmelt backward from a remotely sensed date of disappearance. However, reconstructed SWE estimates are available only retrospectively; they do not provide a forecast. To estimate SWE throughout the snowmelt season, we consider physiographic and remotely sensed information as predictors and reconstructed SWE as the target. The period of analysis matches the AMSR-E radiometer's lifetime from 2003 to 2011, for the months of April through June. The spatial resolution of the predictions is 3.125km, to match the resolution of a microwave brightness temperature product. Two machine learning techniques - bagged regression trees and feed-forward neural networks - produced similar mean results, with 0-14% bias and 46-48mm RMSE on average. Nash-Sutcliffe efficiencies averaged 0.68 for all years. Daily SWE climatology and fractional snow-covered area are the most important predictors. We conclude that these methods can accurately estimate SWE during the snow season in remote mountains, and thereby provide an independent estimate to forecast runoff and validate other methods to assess the snow resource.

Abstract. In many mountains, snowmelt provides most of the runoff. In Afghanistan, few ground-based measurements of the snow resource exist. Operational estimates use imagery from optical and passive microwave sensors, but with their limitations. An accurate approach reconstructs spatially distributed snow water equivalent (SWE) by calculating snowmelt backward from a remotely sensed date of disappearance, but reconstructed SWE estimates are available only retrospectively; they do not provide a forecast. To estimate SWE early in the snowmelt season, we consider physiographic and remotely-sensed information as predictors and reconstructed SWE as the target. The period of analysis matches the AMSR-E radiometer's lifetime from 2003 to 2011, for the months of April through June. The spatial resolution of the predictions is 3.125 km, to match the resolution of a microwave brightness temperature product. Two machine learning techniques – bagged regression trees and feed-forward neural networks – produced similar mean results, with 0–14 % bias and 46–48 mm RMSE on average. Daily SWE climatology and fractional snow-covered area are the most important predictors. We conclude that the methods can accurately estimate SWE during the snow season in remote mountains.

Accurately estimating basin-wide snow water equivalent (SWE) is the most important unsolved problem in mountain hydrology. Models that rely on remotely sensed inputs are especially needed in ranges with few surface measurements. The NASA Airborne Snow Observatory (ASO) provides estimates of SWE at 50 m spatial resolution in several basins across the Western U.S. during the melt season. Primarily, water managers use this information to forecast snowmelt runoff into reservoirs; another impactful use of ASO measurements lies in validating and improving satellite-based snow estimates or models that can scale to whole mountain ranges, even those without ground-based measurements. We compare ASO measurements from 2013 to 2015 to four methods that estimate spatially distributed SWE: two versions of a SWE reconstruction method, spatial interpolation from snow pillows and courses, and NOAA's Snow Data Assimilation System (SNODAS). SWE reconstruction downscales energy forcings to compute potential melt, then multiplies those values by satellite-derived estimates of fractional snow-covered area to calculate snowmelt. The snowpack is then built in reverse from the date the snow is observed to disappear. The two SWE reconstruction models tested include one that employs an energy balance calculation of snowmelt, and one that combines net radiation and degree-day approaches to estimate melt. Our full energy balance model, without ground observations, performed slightly better than spatial interpolation from snow pillows, having no systematic bias and 26% mean absolute error when compared to SWE from ASO. Both reconstruction models and interpolation were more accurate than SNODAS.

Operational ground-based measurements of snow water equivalent (SWE) do not adequately explain spatial variability in mountainous terrain. To address this problem, we combine satellite-based retrievals of fractional snow cover for the period 2000 to 2011 with spatially distributed energy balance calculations to reconstruct SWE values throughout each melt season in the Sierra Nevada of California. Modeled solar radiation, longwave radiation, and air temperature from NLDAS drive the snowmelt model. The modeled solar radiation compares well to ground observations, but modeled longwave radiation is slightly lower than observations. Validation of reconstructed SWE with snow courses and our own snow surveys shows that the model can accurately estimate SWE at the sampled locations in a variety of topographic settings for a range of wet to dry years. The relationships of SWE with elevation and latitude are significantly different for wet, mean and dry years as well as between drainages. In all the basins studied, the relationship between remaining SWE and snow-covered area (SCA) becomes increasingly correlated from March to July as expected because SCA is an important model input. Though the SWE is calculated retrospectively SCA observations are available in near-real time and combined with historical reconstructions may be sufficient for estimating SWE with more confidence as the melt season progresses.

A 11 year data set of spatially distributed snow water equivalent (SWE) was used to inform a quantitative, near-optimal sensor placement methodology for real-time SWE estimation in the American River basin of California. Rank-based clustering was compared with geographically based clustering (subbasin delineation) to determine the existence of stationary covariance structures within the overall SWE data set. The historical SWE data, at 500 × 500 m resolution, were split into 8 years of training and 3 years of validation data. Within each cluster, a quantitative sensor-placement algorithm, based on maximizing the metric of mutual information, was implemented and compared with random placement. Gaussian process models were then built from validation data points selected by the algorithm to evaluate the efficacy of each placement approach. Rank-based clusters remained stable interannually, suggesting that rankings of pixel-by-pixel SWE exhibit stationary features that can be exploited by a sensor-placement algorithm, yielding a 200 mm average root-mean-square error (RMSE) for 20 randomly selected sensing locations. This outperformed geographic and basin-wide placement approaches, which generated 460 and 290 mm RMSE, respectively. Mutual information-based sampling provided the best placement strategy, improving RMSE between 0 and 100 mm compared with random placements. Increasing the number of rank-based clusters consistently lowered average RMSE from 400 mm for 1 cluster to 175 mm for 8 clusters, for 20 total sensors placed. To optimize sensor placement, we recommend a strategy that couples rank-based clustering with mutual information-based sampling design. © 2013. American Geophysical Union. All Rights Reserved.

Snow cover mapping algorithms utilizing multispectral satellite data at various spatial resolutions are available, each treating subpixel variation differently. Past evaluations of snow mapping accuracy typically relied on satellite data collected at a higher spatial resolution than the data in question. However, these optical data cannot characterize snow cover mapping performance under forest canopies or at the meter scale. Here, we use 3 m spatial resolution snow depth maps collected on 116 d by an aerial laser scanner to validate band ratio and spectral-mixture snow cover mapping algorithms. Such a comprehensive evaluation of sub-canopy snow mapping performance has not been undertaken previously. The following standard (produced operationally by an agency) products are evaluated: NASA gap-filled Moderate Resolution Imaging Spectroradiometer (MODIS) MOD10A1F, NASA gap-filled Visible Infrared Imaging Radiometer Suite (VIIRS) VNP10A1F, and United States Geological Survey (USGS) Landsat 8 Level-3 Fractional Snow Covered Area. Two spectral-unmixing approaches are also evaluated: Snow-Covered Area and Grain Size (SCAG) and Snow Property Inversion from Remote Sensing (SPIReS), both of which are gap-filled MODIS products and are also run on Landsat 8. We assess subpixel snow mapping performance while considering the fractional snow-covered area (fSCA), canopy cover, sensor zenith angle, and other variables within six global seasonal snow classes. Metrics are calculated at the pixel and basin scales, including the root-mean-square error (RMSE), bias, and F statistic (a detection measure). The newer MOD10A1F Version 61 and VNP10A1F Version 1 product biases (- 7.1 %, -9.5 %) improve significantly when linear equations developed for older products are applied (2.8 %, -2.7 %) to convert band ratios to fSCA. The F statistics are unchanged (94.4 %, 93.1 %) and the VNP10A1F RMSE improves (18.6 % to 15.7 %), while the MOD10A1F RMSE worsens (12.7 % to 13.7 %). Consistent with previous studies, spectral-unmixing approaches (SCAG, SPIReS) show lower biases (-0.1 %, -0.1 %) and RMSE (12.1 %, 12.0 %), with higher F statistics (95.6 %, 96.1 %) relative to the band ratio approaches for MODIS. Landsat 8 products are all spectral-mixture methods with low biases (-0.4 % to 0.3 %), low RMSE (11.4 % to 15.8 %), and high F statistics (97.3 % to 99.1 %). Spectral-unmixing methods can improve snow cover mapping at the global scale. [ABSTRACT FROM AUTHOR]