Slide 1

By Almoutaz Abdalla Slide 2 Introduction Snow Water Equivalent (SWE) Remote Sensing in Hydrological modeling (snow dominated) Components of Snowmelt modeling Snowmelt Runoff Model (SRM) Slide 3 Microwave radiation emitted from ground, scattered in many directions by the snow grains within the snowpack. Mw emission @ snowpack surface < ground. Factors; snowpack depth & water equivalent, liquid water content, density, grain size and shape, temperature, stratification, snow state, land cover. Slide 4 snow extent, snow depth snow water equivalent snow state (wet/dry) Mw sensitive to snow layer Can be derived Slide 5 A common snowpack measurement instead of depth. Indicates the amount of water contained within the snowpack. Thought as the depth of water that would theoretically result if the entire snowpack melted instantaneously. SWE= Snow Depth (SD) x Snow Density Slide 6 Same depth yields different SWE due to density. SD= 120 density 10% SWE=120 * 0.1= 12 SD= 120 density 40% SWE=120 * 0.4= 48 The density of new snow ranges from about 5% @ T air 14 F, to about 20% @ T air 32 F. After the snow falls its density increases due to gravitational settling, wind packing, melting and recrystallization. Typical values of snow density are 10-20% in the winter and 20-40% in the spring. Slide 7 Same depth yields different SWE due to density. Slide 8 Due to scatterers within the snow pack & increases with thickness and density Snow pack scattering Deeper snow packs Result lower TB TB related to SWE Slide 9 Where, f1 low scattering channel (commonly 18/19 GHz) f2 high scattering channel (commonly 37 GHz) A, B offset and slope of the regression The coefficients should be determined for different climate and land cover conditions. Thus, no single global algorithm can estimate snow depth and/or SWE under all snowpack and land cover conditions. TB related to SWESWE = A+ B [TB(f1,f2) ] Slide 10 Water presence, alters the emissivity of snow, and results higher brightness temperatures For accurate SWE, snow should be in dry conditions. Thus, prefer early morning passes (local time) Slide 11 Depth hoar formation in bottom of the snowpack (mainly in cold regions), will increase the scattering and reduce surface emission, resulting an overestimation of snow depth and SWE. Slide 12 Snow pack changes in time, seasonal aging or metamorphism changes microwave emission of snow. To account for seasonal variability of mw emission from snow, they should be compiled for entire season, over several years. Slide 13 WHY? SWE provides important information for water resources management and is a major research topic in RS assessment of snow cover and melt. Slide 14 Runoff at the outlet of the watershed is an integrated result of the spatially varying sub-parts of the whole basin Similarity is a serious problem in basins with pronounced topography, because of the high spatial variability of hydrometeorological parameters in these regions. Hard/impossible to handle with classical terms Increase in satellite platforms and improvements in the data transmission and processing algorithms, remote sensing (RS) enables to handle these spatial variations. RS is gaining importance in distributed watershed modeling by its spatial variation handling capacity. Slide 15 The advances in RS and GIS enable new data to scientific community. New data necessitate improvements in hydrological modeling rather than using the conventional methods. Existing models are designed for a limited number of types of input and may need to be made more flexible to make optimum use of the range of possible inputs. (Hydalp,2000) New models would enable new input from RS & GIS and provide better hydrological outputs, enabling the understanding of the complex world. Thus, there will be mutual developments between RS, GIS and hydrological modeling, leading improvements in the other ones. Slide 16 Hydrological Modeling Processing Inputs Incoming water Output Discharges Processing vary from black box approaches to physically based models with different degrees of spatial and temporal variability Slide 17 Processing may depend to the type of the precipitation i.e. on rainfall (liquid) or snowfall (solid). When precipitation occurs as snowfall, the discharge timing is not only a function of precipitation timing, but also the heat supplied to the snowpack either by temperature or by radiation. Slide 18 In either forms, the total runoff volume is still total precipitation minus losses; however, snowfall is stored in snowpack until warmer weather allows the phase change from snow/ice to liquid (i.e. melting). Slide 19 Snowmelt runoff modeling has four main components; 1. Extrapolation of meteorological data 2. Point melt rate calculations 3. Integration of melt water over the snow covered areas 4. Runoff Routing Slide 20 Slide 21 In snowmelt dominated basins, very hard if not impossible to find meteorological stations in adequate number and good quality with even distribution. Existing stations mostly located in major valleys rather than the more inaccessible high portions of the basin, where most of the snow exist. Thus, a necessity to use data from a station even though it may be a long distance away and at a much lower elevation from the snowpack. Slide 22 Air temperature, mainly used for two purposes in the snowmelt models. Both as threshold temperature, separating precipitation as rainfall or snow and as critical temperature, used for estimating snowmelt rates. May not to be same and both may be other than zero. Air temperature alters with elevation and temperature lapse rates must be used to convert the measured air temperature at the lower station to the air temperature at the snowpack location. Although, most runoff models assume a fixed value for the lapse rate, the actual value may be a varying value depending on the present meteorological conditions. Often the temperature lapse rate, threshold and critical temperature values are treated as calibration parameters (Hydalp, 2000) of the model used. Slide 23 Distribution of precipitation from point stations to the rest of the basin has been a problem in the hydrology. Come up with, unrealistic and inaccurate results. Besides, the systematic under catch of snow by most rain gauges especially under high wind speeds has long been reported in literature such as Sevruk (1983). Precipitation amount increases with elevation. Slide 24 Exists numerous methods from simple arithmetic averaging, Theissen polygons to inverse distance relations. These methods allow the extrapolation in a horizontal plane (2D), disregarding the topography of the area under study. Some methods such as De-trended Kriging (Garen, 2003) when distributing the meteorological variables takes the topography into consideration. Preliminary study must be performed since some times, distributing the variables in 2D may give better results than distribution in 3D (Weibel et al., 2002). Slide 25 The energy flux that a surface absorbs or emits is dependent to the sum of the: Net all wave radiation (sum of net short (net solar) and long wave (net thermal) radiation) Sensible heat transfer to the surface by turbulent exchange from atmosphere Latent heat of condensation or evaporation Heat added by precipitation (if the temperature of precipitation is different than the surface temperature) Heat conducted from ground Slide 26 Main discrepancy in applying the energy balance is high variability over time and over the location Highly scientifically based equipped automated stations enable the application of the energy based snowmelt models at a point, there still remains the extrapolation of these measurements over the basin. Slide 27 It is rare for such instrumentation to be available at even a single point in a basin, and even then a problem remains in extrapolating the measurements to other parts of the basin (Hydalp, 2000). Slide 28 Instead of measuring all components some approximations are provided, called parametric energy balance methods. In here, some energy components are derived from available data. Such as; Incident radiation, F(latitude, time of year, shading effects, cloud cover and snow albedo). Sensible heat= wind speed * air temperature Precipitation heat supply = Rainfall rate * Rainfall temperature But still, extrapolating over the other parts of the basin ?. And additional assumptions about the seasonal variations of these terms should be made. Slide 29 Air temperature, common factor in all energy balance equations except the net radiation. However, there is generally a good correlation between them (Hydalp, 2000). Temperature can be considered as the driving factor for the day to day variations of the heat supply to snowpack. Slide 30 Depletion of snow cover takes place over a period called melting season during both incident solar radiance and air temperatures increase. SC doesnt disappear at the same time everywhere in the basin. Even uniform melting, differences in initial snow distribution, results variations. (longer SC @ higher elevations) Slide 31 Uncertain runoff predictions may occur even correct melt rates are exist. High SC leads over estimation, low underestimation. Dealing with SC : Two main approaches, Modeled snow pack formation Observed snow pack formation Slide 32 Formation observed; Snow on the surface is monitored. May start with initialization of the melting RS may be helpful Actually one of the practical uses of RS in hydrology since 1970s Basis of Snowmelt Runoff Model (SRM) Slide 33 Formation modeled; Simulation starts at autumn before the melt season. T and P, used to model the snow pack growth and SWE instead of depth used due to compaction of snow SC is given for areas where SWE>0 Ground data useful for cal./val. Slide 34 infiltration percolation Slide 35 Mainly by comparing observed and simulated discharges May also compare SWE and SCA but mostly the hyrographs are compared Slide 36 Goodness of fit (R 2 ) Nash & Sutcliffe Volume Difference Slide 37 Developped by Martinec in 1975 in Swiss Snow and Avalanche Research Institute. Changed & developed with collaboration of Albert Rango (US ARS), Ralph Roberts (US ARS), Michael Baumgartner (University of Bern) Klaus Seidel (University of Zrich) Various versions exist. http: // hydrolab.arsusda.gov/cgibin/srmhome Slide 38 Semi Distributed Semi Physically based Deterministic Same input same output Slide 39 SRM, based on degree day method, can be used to simulate/forecast the snowmelt runoff Simulate daily flows in snowmelt season or year around Provide short term and long term forecasts Analyze the effect of climate change SRM Slide 40 Basin is divided in to elevation zones Precp. & Temp extrapolated from base and snowmelt in each zone computed SCA values are provided to determine melting area information Losses from evap and ground water handled Runoff from all zones summe up before routing Total amount routed by single store Slide 41 Basic snowmelt model Q n+1 = [c Sn. a n (T n + T n ) S n + c Rn. P n ] (A.10000/86400) (1-k n+1 ) + Q n k n+1 Snow melt Rainfall Flow Recession Q : Basin discharge n : Day indicator T : Air temperature P : Precipitation falling as rain S : Snow covered area A : Zonal area k n+1 : Recession coefficient a n : Degree day factor c sn,c rn : correction for losses due to snowmelt and rainfall Slide 42 Basic snowmelt model Q n+1 = [c Sn. a n (T n + T n ) S n + c Rn. P n ] (A.10000/86400) (1-k n+1 ) + Q n k n+1 Snow melt Rainfall Flow Recession uses 7 Parameters3 Variables Slide 43 Variables (Inputs) TemperaturePrecipitation Snow Covered Area % Meteorological Stations Aerial Photos or Satellite Data ForecastedMeasured Slide 44 Variables (Inputs) Slide 45 Variables (Inputs) T & P T or P; either from single station or from separate sites for each zone. Single/synthetic station; T and P lapse rates are needed to extrapolate values. LR can be variable seasonally. P type (snow/rain) f(T crit ) Snow on no SCA temporary snow pack, becomes Q as sufficient melt conditions Rain on snow pack, becomes Q if ripe snow exist Rain on no SCA direct runoff Melting effect of rain is neglected Slide 46 Variables (SCA) Time series of daily SCA, snow depletion curves(SDC) or conventional depletion curves (CDC), is needed Initially ground observations and aerial photos, used. Recently satellite images are utilized. Slide 47 Slide 48 MODIS Slide 49 AQUA & TERRA Or combined product NOAA AVHRR Image Processing Snow covered area determination SCA Slide 50 NOAA/AVHRR Rez 1000 m MODIS Rez 500 m Slide 51 April1/10/12/26 May21/22 June12/13 April1/10/12/26 May22 June11/12/13/ 14/25/30 2004_0892004_137 2004_0972004_145 2004_1052004_153 2004_1132004_161 2004_1212004_169 2004_1292004_177 MODIS MOD10A1 Daily Snow Cover NOAA AVHRR MODIS MOD10A2 8 Daily Snow Cover Slide 52 The discrete points can connected by decreasing gradients linearly. Slide 53 Or by exponential equation. Slide 54 Slide 55 Disappearing patterns expected to be same year to year. Although, differences in winter accumulations and melting conditions may vary. Slide 56 Variables (Inputs) TemperaturePrecipitation Snow Covered Area % Slide 57 7Parameters Runoff Coefficients (c s,c r ) Rainfall Contributing Area Recession Coefficient Degree Day Factor (a) Temperature Lapse Rate ( ) Critical Temperature (T crit ) Time Lag (L) Slide 58 Slide 59 Degree day factor (a) Converts the number of degree days (temperature values above a certain base temperature) (oC d) into snowmelt depths M(cm) M= a*T Comparing the degree day values with the daily decrease of snow water equivalent. Slide 60 Snow pillows, snow lysimeters. Slide 61 In case of no data; can be used. Shows a daily variation, expected as some energy terms are neglected. Slide 62 But, when averaged for a few days, become more stable. Slide 63 As snow ages, snow water content and hence density increases, albedo decreases. All these, favor melting, leading increased ddf. Slide 64 Ddf will maintain its popularity since temperature is tentatively, a good measure of energy flux, in addition to easy to measure and forecast (Martinec and Rango, 1986). Slide 65 Critical Temperature (T crit ) determine the type of precipitation i.e. either rainfall and contribute to runoff immediately (T > Tcrit) or snowfall (T < Tcrit) and lead to accumulation of snowpack and a delayed runoff Thus, new snowfalls are kept in storage until warm days allow the melting. Slide 66 Critical Temperature Tcrit from +3 in April to 0.75 oC in July is reported (WinSRM, 2005) where as +1.5 to 0 oC is reported by US Army Corps of Engineers (1956). Sharp rainfall runoff peaks may be missed by SRM due to the determination of temperature values being less than the Tcrit. Slide 67 Value may be changed, but daily values used, and rain can occur during the warmer or colder period of the day. Slide 68 Temperature Lapse Rate Defines a temperature gradient across the watershed, used in extrapolating temperature values from a given station. SRM accepts a single, basin wide temperature lapse rate or zonal temperature lapse rates. Higher temperature lapse rates for winter and lower values for the summer months are expected (Hydalp, 2000). The depletion of snow cover may represent requirement of the value change of the lapse rate. High temperatures from extrapolation by a LR value but no change in snow areal extent is, then probably no appreciable snowmelt is taking place (WinSRM, 2005) and the LR should be modified accordingly. Slide 69 Slide 70 Runoff Coefficients Explain, differences between the basin runoff and the available precipitation (either snowfall or rainfall) Account for the volume of water, which does not leave the basin, F(the site characteristics, such as soil type, soil depth, elevation, slope, aspect, vegetation type and vegetation density) (Levick, 1998). SRM uses two runoff coefficients c s and c r related to snow melting and rainfall respectively. The two values are expected to be different from each other due to their characteristics. Slide 71 Runoff Coefficients In the early melt period, frozen soil early has lower infiltration and storage capacities. Spring will thaw the soil and snow melt will soak in the soil, leading a drop in the runoff coefficients. As soil becomes saturated, the values will increase again. Thus, monthly variations in runoff coefficients are expected and can be explained by an analysis of the seasonal changes in vegetation and climate (Levick, 1998, Kaya 1999). Slide 72 Slide 73 Time Lag Indicates the time delay between the daily rise in temperature and runoff production. Used for time wise matching of the observed and calculated peaks in the simulation mode. Hydrographs of past years and the daily fluctuating character of the snowmelt enable the predetermination of the time lag value. Value can be modified by comparing the timing of simulated hydrograph peaks with the observed hydrograph peaks. Slide 74 Slide 75 Recession Coefficient Represents the daily melt water production that immediately appears in runoff. Analysis of historic discharge data may be a starting point. Recession from a high discharge is relatively steeper than from a low discharge, which is a commonly observed situation (Seidel and Martinec, 2004). Slide 76 Rainfall Contributing Area If the snow is dry and deep, the snow largely retains the rainfall. Thus, the rainfall directly affecting the runoff values are reduced by the ratio of NO SCA/ Total Area of the zone. As snow softens and ripens, it becomes ready to release the same amount of water as entering to the snowpack. In this case, rainfall falling on the whole area directly affects the hydrograph. The user determines the date of change of the snow condition during the model runs. Slide 77 General Concept of determining SCA MODIS Aqua and Terra Snow Cover Product Image Processing Snow Cover Area Determination Slide 78 Methodology Downloading MODIS daily Terra and Aqua snow cover product Reprojection using MRT tool combine MODIS daily Terra and Aqua snow cover product Classify the combined product - using priority principle Compute the statistics of the snow and cloud percentage Stacked in ENVI and resized to basin area Statistics of stacked images Period from 03/21/2009 to 7/01/2009 Eliminate images that contains more than 20% of cloud cover Derive the snow cover percentage data for the given period Downloading DEM of the area Determining the outlet Delineate the watershed Download the observed Discharge at outlet Download the Snotel data temp, precip. Input the data into snowmelt run off model Running and calibration Compare the simulated discharge with the observed Slide 79 Result images has been stacked in ENVI and resized to the basin area Slide 80 Images that have more than 20% cloud cover Has been eliminated Slide 81 Calibration Cs Coef. Cr Coef. Slide 82 Results R 2 = 0.93 Slide 83 almoutaz@gmail.com