Combining MODIS and AMSR‐E observations to improve MCD43A3 short‐time snow‐covered Albedo estimation

Land surface albedo plays an important role in the radiation budget and global climate models. NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) provide 16‐day albedo product with 500‐m resolution every 8 days (MCD43A3). Some in‐situ albedo measurements were used as the true surface albedo values to validate the MCD43A3 product. As the 16‐day MODIS albedo retrievals do not include snow observations when there is ephemeral snow on the ground surface in a 16‐day period, comparisons between MCD43A3 and 16 day averages of field data do not agree well. Another reason is that the MODIS cannot detect the snow when the area is covered by clouds. The Advanced Microwave Scanning Radiometer for EOS (AMSR‐E) data are not affected by weather conditions and are a good supplement for optical remote sensing in cloudy weather. When the surface is covered by ephemeral snow, the AMSR‐E data can be used as the additional information to retrieve the snow albedo. In this study, we developed an improved method by using the MODIS products and the AMSR‐E snow water equivalent (SWE) product to improve the MCD43A3 short‐time snow‐covered albedo estimation. The MODIS daily snow products MOD10A1 and MYD10A1 both provide snow and cloud information from observations. In our study region, we updated the MODIS daily snow product by combining MOD10A1 and MYD10A1. Then, the product was combined with the AMSR‐E SWE product to generate new daily snow‐cover and SWE products at a spatial resolution of 500 m. New SWE datasets were integrated into the Noah Land Surface Model snow model to calculate the albedo above a snow surface, and these values were then utilized to improve the MODIS 16‐day albedo product. After comparison of the results with in‐situ albedo measurements, we found that the new corrected 16‐day albedo can show the albedo changes during the short snowfall season. For example, from January 25 to March 14, 2007 at the BJ site, the albedo retrieved from snow‐free observations does not indicate the albedo changes affected by snow; the improved albedo conforms well to the in‐situ measurements. The correlation coefficient of the original MODIS albedo and the in‐situ albedo is 0.42 during the ephemeral snow season, but the correlation coefficient of the improved MODIS albedo and the in‐situ albedo is 0.64. It is concluded that the new method is capable of capturing the snow information from AMSR‐E SWE to improve the short‐time snow‐covered albedo estimation. Copyright © 2012 John Wiley & Sons, Ltd.     

Uncertainty of snow water equivalent retrieved from AMSR‐E brightness temperature in northeast Asia

Accurate estimation of snow water equivalent (SWE) has been significantly recognized to improve management and analyses of water resource in specific regions. Although several studies have focused on developing SWE values based on remotely sensed brightness temperatures obtained by microwave sensor systems, it is known that there are still a number of uncertainties in SWE values retrieved from microwave radiometers. Therefore, further research for improving remotely sensed SWE values including global validation should be conducted in unexplored regions such as Northeast Asia. In this regard, we evaluated SWE through comparison of values produced by the Advanced Microwave Scanning Radiometer Earth Observing System (AMSR‐E) from December 2002 to February 2011 with in situ SWE values converted from snow‐depth observation data from four regions in the South Korea. The results from three areas showed similarities which indicated that the AMSR‐E SWE values were overestimated when compared with in situ SWE values, and their Mean Absolute Errors (MAE) by month were relatively small (1.1 to 6.5 mm). Contrariwise, the AMSR‐E SWE values of one area were significantly underestimated when compared with in situ SWE values and the MAE were much greater (4.9 to 35.2 mm). These results were closely related to AMSR‐E algorithm‐related error sources, which we analyzed with respect to topographic characteristics and snow properties. In particular, we found that snow density data used in the AMSR‐E SWE algorithm should be based on reliable in situ data as the current AMSR‐E SWE algorithm cannot reflect the spatio‐temporal variability of snow density values. Additionally, we derived better results considering saturation effect of AMSR‐E SWE. Despite the demise of AMSR‐E, this study's analysis is significant for providing a baseline for the new sensor and suggests parameters important for obtaining more reliable SWE. Copyright © 2013 John Wiley & Sons, Ltd.     

Comparison and validation of MODIS standard and new combination of Terra and Aqua snow cover products in northern Xinjiang,China

Taking northern Xinjiang, China, as an example, this study first compares the standard MODIS Terra and Aqua snow cover classifications, and then compares the accuracy of the standard MODIS daily and 8‐day snow cover products with the new daily and multi‐day snow cover combination of MODIS Terra and Aqua observations using in situ measurements. Under clear sky in both products, the agreement of land classification from MODIS Terra and Aqua daily and 8‐day snow cover products is close to 100% for a entire water year. In contrast, the agreement of snow classification from MODIS Terra and Aqua is high only in the winter months, decreasing in the rest of the period. The high agreement mainly concentrates in land or snow‐dominated areas, and major disagreements take place in the transitions zones from snow to land. The disagreement (mainly snow–land) in the 8‐day products is higher than that in the daily products. In addition, both MODIS Terra and Aqua cloud masks tend to map more areas in the transition zones as cloud. Under clear sky conditions, the three daily products have similar accuracy of snow and land classification, and the 8‐day standard products and the multi‐day combination product also have similar accuracy of snow and land classification. This further suggests that the algorithm in the combination of Terra and Aqua snow cover products is valid. Moreover, in the actual weather/cloud conditions, the combination products from Terra and Aqua reduce cloud blockage and improve snow classification accuracy against either MODIS Terra or Aqua (51% against 44% and 34% for daily and 92% against 87% and 78% for 8‐day, respectively), although Terra snow product (daily or 8‐day) has slightly better accuracy than the Aqua snow product. The new combination products can provide better mapping of spatiotemporal variation of snow cover/glacier and for snow‐melting modeling. Copyright © 2008 John Wiley & Sons, Ltd.     

Influences of forest on MODIS snow cover mapping and snow variations in the Amur River basin in Northeast Asia during 2000–2014

The study applies the improved cloud‐free moderate resolution imaging spectral radiometer daily snow cover product (MODMYD_MC) to investigate the snow cover variations from snow hydrologic year (HY) HY2000 to HY2013 in the Amur River basin (ARB), Northeast Asia. The fractions of forest cover were 38%, 63%, and 47% in 2009 in China (the southern ARB), Russia (the northern ARB), and ARB, respectively. Validation results show that MODMYD_MC has a snow agreement of 88% against in situ snow depth (SD) observations (SD ≥ 4 cm). The agreement is about 10% lower at the forested stations than at the nonforested stations. Snow cover durations (SCDs) from MODMYD_MC are 20 days shorter than ground observations (SD ≥ 1 cm) at the forested stations, whereas they are just 8 days shorter than ground observations (SD ≥ 1 cm) at the nonforested stations. Annual mean SCDs from MODMYD_MC in the forested areas are 21 days shorter than those in the nearby farmland in the Sanjiang Plain. This indicates forest has a complex influence on the snow accumulation and melting processes and even on optical satellite snow cover mapping. Meanwhile, SCD and mean snow cover are negatively correlated with air temperature in ARB, especially in the snow melting season, when mean air temperature in March and April can explain 86% and 74% of the mean snow cover variations in China ARB and Russia ARB, respectively. From 1961 to 2015, the annual mean air temperature presented an increased trend by 0.33 °C/decade in both China ARB and Russia ARB, whereas it had a decrease trend from HY2000 to HY2013. The decrease of air temperature led to an increase of snow cover, which is different from the global decrease trend of snow cover variations. SCD and snow cover had larger increase rates in China ARB than in Russia ARB, and they were larger in the forested areas than in the nearby farmland in the Sanjiang Plain.     

Evaluation of the snow‐covered area data product from MODIS

The Moderate Resolution Imaging Spectroradiometer (MODIS), flown on board the Terra Earth Observing System (EOS) platform launched in December 1999, produces a snow‐covered area (SCA) product. This product is expected to be of better quality than SCA products based on operational satellites (notably GOES and AVHRR), due both to improved spectral resolution and higher spatial resolution of the MODIS instrument. The gridded MODIS SCA product was compared with the SCA product produced and distributed by the National Weather Service National Operational Hydrologic Remote Sensing Center (NOHRSC) for 46 selected days over the Columbia River basin and 32 days over the Missouri River basin during winter and spring of 2000–01. Snow presence or absence was inferred from ground observations of snow depth at 1330 stations in the Missouri River basin and 762 stations in the Columbia River basin, and was compared with the presence/absence classification for the corresponding pixels in the MODIS and NOHRSC SCA products. On average, the MODIS SCA images classified fewer pixels as cloud than NOHRSC, the effect of which was that 15% more of the Columbia basin area could be classified as to presence–absence of snow, while overall there was a statistically insignificant difference over the Missouri basin. Of the pixels classified as cloud free, MODIS misclassified 4% and 5% fewer overall (for the Columbia and Missouri basins respectively) than did the NOHRSC product. When segregated by vegetation cover, forested areas had the greatest differences in fraction of cloud cover reported by the two SCA products, with MODIS classifying 13% and 17% less of the images as cloud for the Missouri and Columbia basins respectively. These differences are particularly important in the Columbia River basin, 39% of which is forested. The ability of MODIS to classify significantly greater amounts of snow in the presence of cloud in more topographically complex, forested, and snow‐dominated areas of these two basins provides valuable information for hydrologic prediction. Copyright © 2003 John Wiley & Sons, Ltd.     

Variability and trends in the annual snow‐cover cycle in Northern Hemisphere land areas, 1972–2000

This study investigated variability and trends in the annual snow‐cover cycle in regions covering high‐latitude and high‐elevation land areas in the Northern Hemisphere. The annual snow‐cover cycle was examined with respect to the week of the last‐observed snow cover in spring (WLS), the week of the first‐observed snow cover in autumn (WFS), and the duration of the snow‐free period (DSF). The analysis used a 29‐year time‐series (1972–2000) of weekly, visible‐band satellite observations of Northern Hemisphere snow cover from NOAA with corrections applied by D. Robinson of Rutgers University Climate Laboratory. Substantial interannual variability was observed in WLS, WFS and DSF (standard deviations of 0·8–1·1, 0·7–0·9 and 1·0–1·4 weeks, respectively), which is related directly to interannual variability in snow‐cover area in the regions and time periods of snow‐cover transition. Over the nearly three‐decade study period, WLS shifted earlier by 3–5 days/decade as determined by linear regression analysis. The observed shifts in the annual snow‐cover cycle underlie a significant trend toward a longer annual snow‐free period. The DSF increased by 5–6 days/decade over the study period, primarily as a result of earlier snow cover disappearance in spring. The observed trends are consistent with reported trends in the timing and length of the active growing season as determined from satellite observations of vegetation greenness and the atmospheric CO₂ record. Copyright © 2002 John Wiley & Sons, Ltd.     

Spatial variability of snowmelt timing from AMSR‐E and SSM/I passive microwave sensors,Pelly River,Yukon Territory,Canada

Spring snow melt run‐off in high latitude and snow‐dominated drainage basins is generally the most significant annual hydrological event. Melt timing, duration, and flow magnitude are highly variable and influence regional climate, geomorphology, and hydrology. Arctic and sub‐arctic regions have sparse long‐term ground observations and these snow‐dominated hydrologic regimes are sensitive to the rapidly warming climate trends that characterize much of the northern latitudes. Passive microwave brightness temperatures are sensitive to changes in the liquid water content of the snow pack and make it possible to detect incipient melt, diurnal melt‐refreeze cycles, and the approximate end of snow cover on the ground over large regions. Special Sensor Microwave Imager (SSM/I) and Advanced Microwave Scanning Radiometer for EOS (AMSR‐E) passive microwave brightness temperatures (T_b) and diurnal amplitude variations (DAV) are used to investigate the spatial variability of snowmelt onset timing (in two stages, ‘DAV onset’ and ‘melt onset’) and duration for a complex sub‐arctic landscape during 2005. The satellites are sensitive to small percentages of liquid water, and therefore represent ‘incipient melt’, a condition somewhat earlier than a traditional definition of a melting snowpack. Incipient melt dates and duration are compared to topography, land cover, and hydrology to investigate the strength and significance of melt timing in heterogeneous landscapes in the Pelly River, a major tributary to the Yukon River. Microwave‐derived melt onset in this region in 2005 occurred from late February to late April. Upland areas melt 1–2 weeks later than lowland areas and have shorter transition periods. Melt timing and duration appear to be influenced by pixel elevation, aspect, and uniformity as well as other factors such as weather and snow mass distribution. The end of the transition season is uniform across sensors and across the basin in spite of a wide variety of pixel characteristics. Copyright © 2007 John Wiley & Sons, Ltd.     

A comparison of snowmelt‐derived streamflow from temperature‐index and modified‐temperature‐index snow models

Reliable estimation of the volume and timing of snowmelt runoff is vital for water supply and flood forecasting in snow‐dominated regions. Snowmelt is often simulated using temperature‐index (TI) models due to their applicability in data‐sparse environments. Previous research has shown that a modified‐TI model, which uses a radiation‐derived proxy temperature instead of air temperature as its surrogate for available energy, can produce more accurate snow‐covered area (SCA) maps than a traditional TI model. However, it is unclear whether the improved SCA maps are associated with improved snow water equivalent (SWE) estimation across the watershed or improved snowmelt‐derived streamflow simulation. This paper evaluates whether a modified‐TI model produces better streamflow estimates than a TI model when they are used within a fully distributed hydrologic model. It further evaluates the performance of the two models when they are calibrated using either point SWE measurements or SCA maps. The Senator Beck Basin in Colorado is used as the study site because its surface is largely bedrock, which reduces the role of infiltration and emphasizes the role of the SWE pattern on streamflow generation. Streamflow is simulated using both models for 6 years. The modified‐TI model produces more accurate streamflow estimates (including flow volume and peak flow rate) than the TI model, likely because the modified‐TI model better reproduces the SWE pattern across the watershed. Both models also produce better performance when calibrated with SCA maps instead of point SWE data, likely because the SCA maps better constrain the space‐time pattern of SWE.