IDENTIFYING AND MAPPING SYSTEMATIC ERRORS IN PASSIVE 
MICROWAVE SNOW WATER EQUIVALENT OBSERVATIONS 
James Foster, 2,3Chaojiao Sun, #Jeffrey P Walker, /.3Richard Kelly, 1,3 Jairui Dong, and I Alfred Chang 
! Hydrological Sciences Branch, Laboratory for Hydrospheric Processes 
NASA Goddard Space Flight Center, Greenbelt, Maryland, 20771 USA 
2Global Modeling and Assimilation Office 
NASA Goddard Space Flight Center, Greenbelt, Maryland, 20771 USA 
3Goddard Earth Sciences and Technology Center 
University of Maryland Baltimore County, Baltimore, Maryland 21250, USA 
4Department of Civil and Environmental Engineering 
University of Melbourne, Parkville, Victoria, 3010 Australia 
KEYWORDS: passive microwave, snowpack, snow crystals, systematic errors, snow water equivalent 
ABSTRACT: 
Understanding remote sensing retrieval errors is important for correct interpretation of observations, and successful 
assimilation of observations into numerical models. Passive microwave sensors onboard satellites can provide global 
snow water equivalent (SWE) observations day or night and under cloudy conditions. However, there are errors 
associated with the passive microwave measurements, which are well known but have not been adequately quantified 
so far. This study proposes a new algorithm for passive microwave SWE retrievals that removes known systematic 
errors. Specifically, we consider the impact of vegetation cover and snow crystal growth on passive microwave 
responses. As a case study, systematic errors (difference between the old and new algorithms) are presented for the 
snow season 1990-91. Standard error propagation theory is used to estimate the uncertainty in the new retrieval 
algorithm (not shown here). An unbiased SWE dataset is produced and monthly SWE error maps (October-May) are 
derived for the Northern Hemisphere. The next step is to fine tune and test the bias-free algorithm, which will be 
applied to the combined passive microwave dataset from SMMR and SSM/I over 20 years. 
I. INTRODUCTION 
Snow plays an important role in the global energy and 
water budgets, as a result of its high albedo and 
thermal and water storage properties. Snow is also the 
largest varying landscape feature of the Earth's 
surface. Thus, knowledge of snow extent and SWE are 
important for climate change studies and applications 
such as flood forecasting. Furthermore, snow depth 
and SWE, as well as snow cover extent, are important 
contributors to both local and remote climate systems. 
Despite its importance, the successful forecasting of 
snowmelt using atmospheric and hydrologic models is 
challenging. This is due to imperfect knowledge of 
snow physics and simplifications used in the model, as 
well as errors in the model forcing data. Furthermore, 
the natural spatial and temporal variability of snow 
cover is characterized at space and time scales below 
those typically represented by models. Snow model 
initialization based on model spin-up will be affected 
by these errors. By assimilating snow observation 
products into Land Surface Models (LSMs), the 
effects of model initialization error may be reduced 
(Sun et al., 2003). 
Passive microwave remote sensors onboard satellites 
provide an all-weather global SWE observation 
capability day or night. Brightness temperatures from 
different channels of satellite passive microwave 
sensors (hereafter referred to as PM) can be used to 
estimate the snow water equivalent (or snow depth 
with knowledge of the snow density), and hence snow 
cover extent. However, there are both systematic 
(bias) and random errors associated with the passive 
microwave measurements. In order for the remotely 
sensed SWE observations to be useful for climate 
modelers, water resource managers and flood 
forecasters, it is necessary to have both an unbiased 
SWE estimate and a quantitative, rather than 
qualitative, estimate of the uncertainty. This is a 
critical requirement for successful assimilation. of 
snow observations into LSMs. 
For most PM algorithms, the effects of vegetation 
cover and snow grain size variability are the main 
source of error in estimating SWE. Of lesser concern 
are the effects of topography and atmospheric 
conditions. A major assumption made in a number of 
PM algorithms is that vegetation cover does not affect 
the SWE estimates. In fact, it can have a significant 
impact on the accuracy of SWE estimates. In densely 
forested areas, such as the boreal forest of Canada, the 
underestimation of SWE from retrieval algorithms can 
be as high as 5096 (Chang et al., 1996). Another major 
assumption is that snow density and snow crystal size 
remain constant throughout the snow season 
everywhere on the globe; in reality, they vary 
considerably over time and space. The PM algorithms 
     
   
   
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
   
    
   
  
  
  
  
  
     
   
    
   
  
  
  
  
  
  
  
     
  
    
Intern 
—Z 
are 
(F 
Th 
for 
ob 
ass 
pa 
an 
(S 
mi 
GI 
we Un mo" CI o eS