Quantifying the uncertainty in passive microwave snow water equivalent observations

James L. Foster, Chaojiao Sun, Jeffrey P. Walker, Richard Kelly, Alfred Chang, Jiarui Dong, Hugh Powell
2005 Remote Sensing of Environment  
14 Passive microwave sensors (PM) onboard satellites have the capability to provide global snow observations which are not affected by 15 cloudiness and night condition (except when precipitating events are occurring). Furthermore, they provide information on snow mass, i.e., 16 snow water equivalent (SWE), which is critically important for hydrological modeling and water resource management. However, the errors 17 associated with the passive microwave measurements of snow water equivalent are
more » ... ell known but have not been adequately quantified thus 18 far. Understanding these errors is important for correct interpretation of remotely sensed SWE and successful assimilation of such 19 observations into numerical models. 20 This study uses a novel approach to quantify these errors by taking into account various factors that impact passive microwave responses 21 from snow in various climatic/geographic regions. Among these factors are vegetation cover (particularly forest cover), snow morphology 22 (crystal size), and errors related to brightness temperature calibration. A time-evolving retrieval algorithm that considers the evolution of 23 snow crystals is formulated. An error model is developed based on the standard error estimation theory. This new algorithm and error 24 estimation method is applied to the passive microwave data from Special Sensor Microwave/Imager (SSM/I) during the 1990-1991 snow 25 season to produce annotated error maps for North America. The algorithm has been validated for seven snow seasons (from 1988 to 1995) in 26 taiga, tundra, alpine, prairie, and maritime regions of Canada using in situ SWE data from the Meteorological Service of Canada (MSC) and 27 satellite passive microwave observations. An ongoing study is applying this methodology to passive microwave measurements from 28 Scanning Multichannel Microwave Radiometer (SMMR); future study will further refine and extend the analysis globally, and produce an 29 improved SWE dataset using the new algorithm for North America over 20 years by combining SSMR and SSM/I measurements. 30 Published by Elsevier Inc. 31 (J.L. Foster). Remote Sensing of Environment xx (2004) xxx -xxx www.elsevier.com/locate/rse RSE-06206; No of Pages 18 DTD 5 ARTICLE IN PRESS 46 extent and SWE are important for climate change studies 47 and applications such as flood forecasting. 48 Despite its importance, the successful forecasting of 49 snowmelt using atmospheric and hydrologic models is 50 challenging. This is due to the imperfect knowledge of 51 snow physics and simplifications used in the model, as well 52 as errors in the model forcing data. Furthermore, the natural 53 spatial and temporal variability of snow cover is charac-54 terized at space and time scales below those typically 55 represented by models. Snow model initialization based on 56 model spin-up will be affected by these errors. By 57 assimilating snow observation products into land surface 58 models, the effects of model initialization error may be 59 reduced (Sun et al., 2004) . 60 A critical requirement for successful assimilation of snow 61 observations into models is an accurate knowledge of the 62 observation errors. While it is possible to directly replace 63 modeled states with observed states, this does not take into 64 account the fact that model predictions and remotely sensed 65 observations contain different amounts of error. In state-of-66 art data assimilation, error statistics of the observational data 67 are required so that the correct weighting between observa-68 tions and model estimates may be applied. Furthermore, in 69 order for the remotely sensed SWE observations to be useful 70 for climate modelers, water resource managers, and flood 71 forecasters, it is necessary to have a quantitative, rather than 72 qualitative, estimate of the uncertainty. A framework is 73 needed to estimate SWE and its associated errors over large 74 geographic areas. 75 In situ SWE data are poorly distributed globally and 76 collected irregularly (Robinson et al., 1993) . Passive micro-77 wave remote sensors onboard satellites provide an all-78 weather global SWE observation capability day or night. 79 Brightness temperatures from different channels of satellite 80 passive microwave sensors (hereafter referred to as PM) can 81 be used to estimate SWE (or snow depth with knowledge of 82 the snow density), and hence snow cover extent. This is a 83 significant advantage over infrared sensors, which only work 84 under cloud-free conditions, and visible sensors, which also 85 require daylight to observe terrestrial features. More impor-86 tantly, PM sensors provide estimates of the snow mass and 87 not just snow cover extent. However, there are errors 88 associated with the PM measurements. In order for the 89 remotely sensed SWE observations to be useful for climate 90 modelers, water resource managers, and flood forecasters, it 91 is necessary to have both an unbiased SWE estimate and a 92 quantitative, rather than qualitative, estimate of the uncer-93 tainty. This is a critical requirement for successful assim-94 ilation of snow observations into land surface models. 95 For most PM algorithms, the effects of vegetation cover 96 and snow grain size variability are the main source of error 97 in estimating SWE. Of lesser concern are the effects of 98 topography and atmospheric conditions. A major assump-99 tion made in a number of PM algorithms is that vegetation 100 cover does not affect the SWE estimates. In fact, it can have 101 a significant impact on the accuracy of SWE estimates. In
doi:10.1016/j.rse.2004.09.012 fatcat:u7ejc4zcyzerxoaadljffgpi6a