An ensemble Kalman filter is used to assimilate 1 Advanced Microwave Scanning Radiometer-2 (AMSR2) obser-2 vations of passive microwave (PMW) brightness temperatures 3 (spectral differences, ∆T b ) into land surface model estimates 4 of snow mass over northwestern high mountain Asia (HMA). 5 Trained support vector machines (SVMs) serve as the observa-6 tion operator and map the geophysical modeled variables into 7 ∆T b space within the assimilation framework. Evaluation of 8 the assimilation

more »

... tine is carried out through comparison of 9 assimilated snow mass estimates with an in situ dataset. The 10 assimilation framework helps improve the land surface model 11 estimates through PMW ∆T b assimilation, particularly in terms 12 of decreasing the domain-wide bias. The assimilation framework 13 proved more effective during the (dry) snow accumulation season 14 and decreased the bias and RMSE in snow mass estimates at 15 76% and 58% of the comparative pixels, respectively. During the 16 snow ablation season, the PMW brightness temperature signal 17 contained less information related to snow mass due to the 18 presence of other concurrent geophysical features that effectively 19 serve as noise during the snow mass update. The utilization of 20 PMW ∆T b for accurate snow mass estimation in complex terrain 21 such as HMA is dependent on a multitude of factors for optimal 22 results; however, it does add utility to the land surface model 23 if the relevant pitfalls are taken into consideration prior to the 24 state variable update. 25 Index Terms-passive microwave, brightness temperature, land 26 surface modeling, hydrology, snow, high mountain Asia, NASA 27 Land Information System 28 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. applied over the whole data-scarce HMA domain, including 85 the Hindu Kush and Himalaya mountain ranges in the west and 86 center as well as the monsoon-dominated regions in the center 87 and east of HMA. In this study, we explore the utilization of 88 data assimilation to formulate a consistent technique for SWE 89 estimation over the whole HMA domain. 90 Data assimilation (DA) is the integration of observed data 91 into model estimates. Synthetic experiments carried out by 92 Kwon et al. [14] to study the applicability of PMW ∆T b 93 assimilation to improve SWE estimation across HMA showed 94 that the assimilation framework was effective in improving 95 SWE estimates for SWE depths < 200 mm during dry 96 snowpack conditions. Building upon these synthetic tests, this 97 study utilizes real-world PMW ∆T b observations from the 98 Advanced Microwave Scanning Radiometer-2 (AMSR2) in an 99 attempt to improve snow mass estimates across HMA. 100 II. STUDY DOMAIN 101 High mountain Asia (HMA) is generally defined as the high 102 elevation region within the Asian continent that spans over 103 eight countries-Tajikistan, 104 Nepal, Bhutan, and Bangladesh-and three main mountain 105 ranges-Hindu Kush, Karakorum, and Himalaya (Fig. 1). 106 According to the Sturm and Holmgren classification, the 107 snow in HMA is primarily composed of prairie and ephemeral 108 snow types along with the presence of alpine and maritime 109 snow near the glacier zones located at the border between 110 Pakistan and China [15]. Hammond et al. [16] observed that 111 low snow zones coincide with areas of low elevation and 112 that snow persistence increases with elevation. Several studies 113 have analyzed the loss of snow cover and glacier melt under 114 evolving climatic scenarios in this region [17]-[19]. Hetero-115 geneous trends in seasonal SWE were reported across HMA 116 based on a PMW-based analysis of change in seasonal SWE 117 [20]. Changes in seasonal snow affect the downstream runoff, 118 especially for the Indus Basin [21]. Cryospheric monitoring 119 of this area is as important as it is difficult due to the harsh 120 climate and inaccessibility of the mountainous regions. 121 Remote sensing in HMA is a complex process primarily 122 due to the high spatial variability in elevation, relatively coarse 123 resolution of available satellite data, the relatively consistent 124 presence of clouds, and a general lack of ground-based mea-125 surements for model validation and evaluation purposes. In 126 such situations, data assimilation (DA) aids in understanding 127 and improving the estimation of the various geophysical 128 variables such as SWE. In this study, the northwestern part 129 of HMA spanning Pakistan, Afghanistan, and Tajikistan is 130 examined. Data assimilation (DA) of PMW ∆T b is attempted 131 to improve snow mass estimates in this region. Further details 132 regarding the data assimilation framework are provided in 133 Section III, the results of which employ in situ data during 134 evaluation of the DA results. 135