1: \section{Modeling approach}
2:
3: \newcommand{\degree}{\mbox{$^{\circ}$}}
4: The core of the model is a spectrum-resolving (line-by-line)
5: atmospheric/surface radiative transfer model, SMART, (Meadows and Crisp, 1996; Crisp, 1997). AIRS (Atmospheric
6: Infrared Sounder) data were used as input for Earth surface albedos and atmospheric properties (Fishbein et al., 2003).
7:
8:
9:
10: These
11: data were processed and mapped on the sphere using a pixelization
12: scheme created for Planck and WMAP missions (Healpix) (Gorski et al., 1998)
13: (fig. \ref{fig:healpix}). For each of the
14: pixels SMART was run to generate a database of synthetic spectra for a
15: variety of viewing angles, illuminations, surface types and cloud cover.
16: Finally, a model was created in which the user specifies spatial resolution
17: and observing position, and the program selects the appropriate synthetic
18: spectra and creates the final view, which can then be
19: averaged on the disk (Tinetti et al, 2004).
20:
21: In the following paragraphs we give a
22: fairly detailed description of the models, of the data we used as input to model,
23: and of the methods we have selected to estimate some specific parameters or
24: indexes.
25:
26:
27:
28: \begin{figure}[h!]
29: \begin{center}
30: \includegraphics[width=11cm]{Resolution_Needed_to_Resolve_Disk.eps}
31: \caption{ { \footnotesize \emph{
32: For producing disk-averages that could be viewed from any direction, spherical coordinates were not the best choice:
33: parallels and meridians do not divide the sphere into equal area
34: tiles, the pole is a singularity, the geometry is not invariant for a
35: general rotation.
36: \newline
37: We chose to use Healpix (Hierarchical Equal
38: Area and iso-latitude Pixelization) grid as tiling program for our
39: model. Healpix, originally designed for the ESA Planck and
40: NASA WMAP missions (Gorski et al., 1998), is a curvilinear partition of the sphere into exactly
41: equal area quadrilaterals . This peculiarity is particularly desirable
42: since the geometry does not considerably change when we look at the planet
43: from different viewing angles, and the contribution of each region of
44: the sphere is accounted for with the same weight. The ability to
45: easily degrade and upgrade the resolution of the map, is also
46: useful.
47: \newline
48: We resolved the Earth with 48 pixels for mapping the atmospheric
49: properties and more than three thousands pixels for tiling the earth's
50: surface.
51: } } } \label{fig:healpix}
52: \end{center}
53: \end{figure}
54:
55:
56:
57: \subsection{Input data}
58: \paragraph{Atmospheric properties}
59: The atmospheric temperature, water vapor trace-gas and cloud
60: distributions needed to construct the model are extracted from the
61: NCEP Global Forecast System (NCEP, 1988; Kanamitsu et al, 1991;
62: Kanamitsu, 1989) and supplemented by the Upper Atmospheric Research
63: Satellite (UARS) and Harvard University tropospheric ozone
64: climatologies by the AIRS Level~2 Simulation System (Fishbein et al.,
65: 2003). The NCEP forecasts are produced
66: every three hours and have 1\degree times 1\degree horizontal
67: resolution and extend upto 10\,hPa. The UARS climatologies
68: are monthly zonal means with 10\degree latitude resolution and extent
69: above 0.01\,hPa while the Harvard tropospheric ozone climatology
70: (Logan, 1998) is an annual mean with 4\degree times 5\degree resolution.
71:
72: % I believe this paragraph can be deleted - this detail is irrelevant
73: % for the result.
74:
75: The simulation system samples the fields as the AIRS instrument would
76: view the earth from its 705\,km altitude sun-synchronous 1:30\,PM
77: local-time ascending equator-crossing polar orbit. The satellite
78: orbits the earth 14.6 times per day collecting 3 million observations
79: daily. The AIRS scan produces gores between the orbit passes near the
80: equator (Fig. \ref{fig:stempday}) which have been filled using a
81: Gaussian interpolation. The data have been remapped to a spherical
82: grid, 37 by 72 (latitude by longitude) (Fig. \ref{fig:stempday}).
83:
84: \begin{figure}[h!]
85: \includegraphics[width=7cm]{albedobefore.eps}
86: \includegraphics[width=7cm]{albedoafter.eps}
87: \caption{ { \footnotesize \emph{ Left: This figure shows the actual
88: AIRS solar albedo map. The dark blue strips are the regions
89: where there are no data points. Right: This figure shows the
90: same data after using the two-dimensional Gaussian
91: interpolation. The strips are no longer visible but the
92: contours are less defined. \newline
93: The spatial interpolation method assigned each of the 3 million data
94: points (taken along the near polar orbit path) to a regular latitude-longitude grid by performing a weighted average, using a two-dimensional
95: Gaussian distribution centered at the location of the data point. The average value at each grid
96: point was normalized by summing the contributions of 3 million
97: Gaussians at each latitude/longitude grid point. This procedure was repeated for each parameter -
98: temperature and trace gases. } } } \label{fig:stempday}
99: \end{figure}
100: % This can also probably be deleted.
101: %
102:
103: \begin{figure}[h!]
104: \includegraphics[width=7cm,angle=0]{pt.eps}
105: \includegraphics[width=7cm,angle=0]{atm.eps}
106: \caption{ { \footnotesize \emph{
107: Examples of temperature-pressure (fig. on the left) and gas-mixing ratios profiles
108: (H$_{2}$O, O$_{2}$, O$_{3}$, CH$_{4}$,
109: CO$_{2}$, CO, N$_{2}$O, fig. on the right) used for our model
110: } } } \label{fig:atm}
111: \end{figure}
112:
113:
114:
115:
116: %To ensure that the area represented by a data point is substantially
117: %less than the area of a base resolution pixel, the existing 37 by 72
118: %grid was first interpolated to a very fine resolution
119: %grid %(we have arbitrarily chosen a 2160 by 2880 grid)
120: %Next, each point was assigned to one of the 48 pixels based on its
121: %position on the earth's surface and the data values were averaged over
122: %each pixel. %based on the number of data points contained in that pixel.
123: %The program resulted in the generation of 48 sets of data, one for
124: %each pixel. Each set had temperature values and trace-gas volume
125: %mixing ratios for 50 pressure levels. To avoid the overemphasis of
126: %the data points near the poles, each point on the high-resolution grid
127: %was area-weighted (a point near the equator was given more weight than
128: %a point near the poles).
129:
130: \paragraph{Solar spectrum}
131:
132: The upper boundary condition of the model is specified by the downward solar flux at the top of the atmosphere. The solar spectrum used is a high-spectral-resolution spectrum compiled from satellite observations at UV wavelengths (Atlas-2 SUSIM) and stellar atmosphere models at longer wavelengths
133: (Kurucz, 1995). The position of the sun in our simulation was derived from the Jet Propulsion Laboratory Horizons Ephemeris System (Giorgini et al., 1997). \\
134:
135:
136:
137:
138: \paragraph{Molecular Line lists}
139:
140: The optical properties of the atmosphere were simulated using molecular line databases to generate the line-by-line description of infrared vibration-rotation bands of gases, and UV and visible cross-sections of gases associated with electronic and pre-dissociation transitions (e.g. ozone).
141:
142: For our model, we have used the HITRAN2K
143: (High resolution Transmission 2K) database [Rothman et al., 2003].
144:
145: \paragraph{Albedo/Surface Types}
146: The land cover map provided by AIRS assigns to each point one of
147: eighteen possible surface types (Fig. \ref{fig:ndvi}). The
148: contribution of each material is determined by land fraction, the
149: amount of vegetation, and the types of vegetation defined by the
150: International Geosphere Biosphere Programme (IGBP) land use surface
151: classification (Loveland and Belward, 1997). For our purposes, these 18 divisions were
152: then reduced to six general surface types: ocean,
153: forest, grass, tundra, ground, and ice.
154:
155:
156: \begin{figure}[h!]
157: \begin{center}
158: \includegraphics[width=14cm,bbllx=0bp,bblly=180bp,bburx=608bp,bbury=626bp,clip=]
159: {ndvi.eps}
160: \caption{ { \footnotesize \emph{ The IGBP Global Cover Type map)
161: \newline
162: } } } \label{fig:ndvi}
163: \end{center}
164: \end{figure}
165:
166: These different surface types were simulated using as input to the
167: radiative transfer model six measured, wavelength dependent, surface
168: reflectance spectra, $\alpha_{k} (\lambda), \, k = 1 \ldots 6$.
169:
170: \begin{figure}[h!]
171: \begin{center}
172: \includegraphics[width=8cm,angle=0]{alb.eps}
173: % \includegraphics[width=8cm]{albedoafter.eps}
174: \caption{ { \footnotesize \emph{ Reflectance in the optical/NIR for
175: the six surface types we used in our model } } }
176: \label{fig:albedo}
177: \end{center}
178: \end{figure}
179:
180: The IGBP map (Fig. \ref{fig:ndvi}) and the solar albedo map (fig.
181: \ref{fig:stempday}) must be consistent. The surface reflectance
182: spectra $\alpha_{k} (\lambda)$ must then be recalibrated in order
183: to give finally the same global average albedo provided by the AIRS
184: albedo map. The scaling factor for each surface type can be calculated
185: by comparing the solar albedo value from the AIRS map (fig.
186: \ref{fig:stempday}) with the numerical integral in the 0.49-0.94
187: $\mu$m range (the wavelength range of the AIRS instrument which
188: provided the albedo map) of the $\alpha_{k} (\lambda)$. Using the
189: simplified land cover map (6 surface types only), for each of the 48
190: atmospheric pixels, the fractional area covered by each of the six
191: surface types was determined. The radiative transfer code was run only
192: for the surface types with a non-negligible fractional area (greater
193: than 5\% of the entire pixel area).
194:
195: %\subsection*{Methods}
196:
197:
198:
199:
200: \paragraph{Clouds}
201: Clouds vary considerably in amount and type over the globe. They have very important effects on longwave and solar energy transfer in the atmosphere, as we can see from Table 1 (Harrison et al. 1990), (D. Hartmann, 1994), (Oort and Peixoto, 1992).
202:
203: %It is possible to measure the radiative fluxes of energy entering and leaving Earth from satellites.
204:
205: %\begin{tabular}[h!]{lccc}
206: % & Average & cloud-free & cloud forcing \\
207: %Outgoing longwave radiation & 234 & 266 & +31 \\
208: %Absorbed solar radiation & 239 & 288 & -48 \\
209: %Net radiation & +5 & +22 & -17 \\
210: %Albedo & 30\% & 15\% & +15\% \\
211: %\end{tabular} \\ {\par \footnotesize{ \textbf{Table 1 -} Radiative flux densities are given in W m$^{-2}$ and albedo in percent. From (Harrison et al. 1990)} } \\
212:
213: For radiative calculations optical and radiative properties must be specified. Normally, water clouds have relatively weak solar absorption, but they effectively scatter solar radiation back toward space and so have high reflectivities. Thick clouds behave almost like black bodies for long wave radiation: they absorb all incident long wave radiation and emit like a blackbody with the temperature of the atmosphere at that the same level of the cloud. A simple approach is to specify the properties of three types of clouds: high (cirrus), medium (cumulus) and low (stratus). \\ \\
214:
215: \begin{tabular}[h!]{lccc}
216: Type & SW reflectivity & SW absorptivity & \% area \\
217: High (cirrus) & 0.21 & 0.005 & 0.228 \\
218: Medium (cumulus) & 0.48 & 0.02 & 0.09 \\
219: Low (stratus) & 0.69 & 0.035 & 0.313 \\
220: \end{tabular}
221: \\
222: {\par \footnotesize{ \textbf{Table 1 -} Values of cloud shortwave reflectivity and absorptivity and fractional area coverage assumed in (Manabe and Strickler, 1964)} } \\
223:
224: In our model, the wavelength dependent optical properties of liquid water
225: droplets were computed using Mie scattering theory (Crisp, 1997), (Liou, 2002).
226: Cirrus clouds were parameterized as polydispersions of hexagonal crystals, and their optical properties were derived using geometric optics
227: (Muinonen et al., 1989). The liquid water and ice refractive indexes were obtained from (Segelstein,1981) and (Warren, 1984) respectively.
228:
229: The top and bottom pressures and the optical depths were taken from AIRS Level~2 Simulation System (Fishbein et al.,
230: 2003). Those sets of data are sometimes underestimating the presence of low clouds.
231: We have used both AIRS and MODIS data (MODIS website) for the cloud maps.
232:
233:
234: %According to the observations, high clouds are concentrated in the convection zones of the tropics over the equatorial South America and South Africa, and a major concentration exists over Indonesia/eastern India and western Pacific oceans. Low clouds are most prevalent in the subtropical eastern ocean margins and in middle latitudes. The low cloud concentrations in the eastern subtropical are explained with lower than average sea surface temperature and consist of strato-cumulus clouds trapped below an inversion. Low clouds are heavily concentrated over the oceanic regions and are less commonly observed over land. Minima in total cloud cover occur in the subtropics in desert regions, over the Caribbean Sea, and southern subtropical zones of the Pacific, Atlantic and Indian Ocean.
235:
236:
237:
238:
239:
240:
241:
242:
243:
244:
245:
246: