MAIN PAGE

MODIS INFO

INSTRUMENT OPS
 Weekly Reports
 MODIS History

CALIBRATION

L1B PRODUCT INFO
& STATUS

CONTACTS & STAFF

ON-LINE DOCS

FREQUENTLY
ASKED QUESTIONS

SITE MAP
SEARCH
RELATED LINKS
MCST DICTIONARY

NASA/GSFC Website Privacy and Security Statement, Disclaimer, and Accessibility Certification

MCST Frequently Asked Questions

Questions

1. What wavelength does each MODIS band detect?
2. How can scaled integers (SI) be converted to radiances and reflectances?
3. How can I extract temperature from the L1B data product?
4. Where can I find Relative Spectral Response (RSR) data?
5. Why are Bands 13 and 14 listed as high and low gain?
6. Where can I find the MODIS black-body temperature?
7. Why are the scans not lined up even at nadir? How can I correct for this effect?
8. Why are some reflectance values negative? Should I disregard those pixels?
9. Why are Band 26 values taken from EV_1KM_RefSB different from those in EV_Band26?
10. How are the MODIS counts converted to radiances and reflectances?
11. Can we correct our land imagery to a common solar zenith angle?
12. Images seem distorted at the edges of scans. Why, and what can be done?
13. Why do white spots occur on the MODIS L1B images?

Answers

What wavelength does each MODIS band detect?

MODIS band specifications are listed at the MODIS Technical Specifications page on the main MODIS web site.

How can scaled integers (SI) be converted to radiances and reflectances?

Level 1B software converts the geo-located raw instrument data numbers (DN) into top of the atmosphere (TOA) calibrated radiances for all bands and into reflectances for the 20 reflected solar bands (RSB).

The L1B values reported for each band/pixel are 'scaled integers' (SI).  Keep in mind that valid SI values are in the range 0 to 32767. Values greater than 32767 are invalid “fill” values, that indicate missing pixels or data that cannot be calibrated for various reasons.

The conversion from SI to radiances and reflectances is described in the MODIS L1B Product User's Guide, available at MCST's web site: http://www.mcst.ssai.biz/mcstweb/L1B/product.html

SI can be converted to radiance values using the band-dependent 'radiance_scales' and radiance_offset' values found in the hdf file.  To convert the 1KM Emissive Bands Scaled Integers to radiances, convert the SI into the radiance values using
radiance = rad_scales*(SI-rad_offset).

The reflectance_scales and reflectance_offsets values are located in the L1B HDF's Vgroup, Data Fields, and in the Dataset Attributes. These values are band dependent.

How can I extract temperature from the L1B data product?

The L1B values reported for each band/pixel are 'scaled integers' (SI). These can be converted to radiance values using the band-dependent 'radiance_scales' and radiance_offset' values including in the hdf file. In turn the radiance can be converted to brightness temperature (top-of-atmosphere) using the Planck function.

It is not possible to derive temperature from L1B radiances without knowing the emissivity of the observed object, or by performing inter-band comparisons. However, *brightness temperature* (the temperature for an ideal blackbody with the observed radiance) may be calculated from Planck's Law, eJuly 3, 2008g src="../images/planck.gif" width="107" height="72" />

   where

   L = radiance (Watts/m2/steradian/m)
   h = Planck's constant (joule second)
   c = speed of light in vacuum (m/s)
   k = Boltzmann gas constant (joules/Kelvin)
   l = band or detector center wavelength (m)
   T = temperature (Kelvin)

Inverting this formula to solve for temperature gives

For a more precise conversion, you may wish to apply the filter response functions (aka Relative Spectral Response) to apply each detector's wavelength response individually, and to weight the calculated temperature at each wavelength interval to get a centroid brightness temperature.

<http://www.mcst.ssai.biz/mcstweb/L1B/product.html>

To derive surface temperature you will also need to consider corrections for atmospheric contributions to the satellite-received signal.

Where can I find Relative Spectral Response (RSR) data?

Current RSR files can be found on our ftp site, for both the Terra-MODIS (PFM) and Aqua-MODIS (FM1) instruments. The RSR tables are in separate files for each band; please see the Readme files for format information.

Why are Bands 13 and 14 listed as high and low gain?

Bands 13 and 14 are each set up to provide two different sensitivities with the same bandpass, to satisfy different scientific needs. In order to maintain acceptable signal-to-noise ratio, each band consists of two parallel arrays of 10 identical detectors. The "Time Delay Integration" (TDI) circuit provides a precise delay between the signal for two adjacent detectors so that the signal from each detector samples the same geography. The two signals are added, then split and put through two different amplifications.

The signal sequence can be thought of as follows:

• The first detector receives light from a 1-km square area on the earth's surface.
• One frame (333µs) later, the scan mirror has rotated to illuminate the adjacent detector with light from the same area
• The TDI circuit delays the first signal by exactly 333µs, and adds the two signals.
• The combined analog signal is split into two paths.
• A different gain is applied to each signal voltage path.
• Each analog signal is digitized.
• Each digital signal is read as a separate band.

Note that what we sometimes call a "detector" in Bands 13 and 14, high and low, would be more properly called a "channel" to distinguish signal path from physical chunks of silicon.

Where can I find the MODIS black-body temperature?

The MODIS black-body temperature data exist in raw DN form, in both the Level 1A file and the Level 1B OBC file. File specifications for these products may be found on the Level 1B Product Information and Status page.

Each file contains a Vdata named "Engineering BB data". There are 12 fields corresponding to the 12 BB thermistors. The field names are "TP_BB_TEMP01", ..., "TP_BB_TEMP12". The formulas implemented in Level 1B code to convert the telemetry DNs to temperatures in engineering units are from a memo written by Jack Xiong and Tim Dorman of MCST.

In addition to the raw DNs, the granule average BB temperature in Kelvin is reported in each of the four L1B products in the ECS core metadata as "AveragedBlackBodyTemperature", field "PARAMETERVALUE.1".

Why are the scans not lined up even at nadir? How can I correct for this effect?

Each successive scan of a MODIS Level 1B image appears to be shifted Westward due to the Earth's rotation during the 1.48 seconds between each scan. This effect will be greatest near the equator (about 1km), and diminishes toward the poles.

To correct for this effect, you must use the geolocation data provided in MOD02*KM Science Data Sets "Latitude" and "Longitude" to map each pixel onto a Lat/Lon grid. One difficulty is that these coordinates are provided only to 1KM resolution for the 250m data product; if you want to map the full resolution, you must interpolate/extrapolate these coordinates to every pixel of each scan.

Why are some reflectance values negative? Should I disregard those pixels?

For reflective bands the corrected signal (dn**) is scaled to a range of 0 to 32767, and reflectance or radiance values are recovered by applying the appropriate calibration coefficients. For some reflective bands (currently 8 - 19 and 26), negative radiances are recorded in order to preserve all potentially useful data.

Each pixel has an assigned Uncertainty Index (UI), which is usually scalable to uncertainty in percent of reflectance. If a pixel is unusable, the UI will be set to 15; in this case the calculated uncertainty will not be correct. For each pixel having negative reflectance, the UI is set to 15.

In brief, the Uncertainty Index should be taken into account when evaluating data; pixels with a UI of 15 should not be used.

Why are Band 26 values taken from SDS EV_1KM_RefSB different from those in SDS EV_Band26?

Data is put into the Science Data Sets on a scan-by-scan basis. When a scan is in "night" mode, Band 26 data is not filled into the EV_1KM_RefSB data set - but is filled into the EV_Band26 data set. If you want all possible Band 26 data, be sure to extract values from SDS EV_Band26, since night mode granules and mixed day/night granules will have different values in the two data sets.

How are the MODIS counts converted to radiances and reflectances?

The process is described in the MODIS L1B Product User's Guide at MCST's web site:

http://www.mcst.ssai.biz/mcstweb/documents/L1B_Product_Users_Guide.pdf

Can we correct our land imagery to common solar zenith angle ?

To date we are looking at the MCD43B4 data and corresponding BRDF parameters held in MCD43B1 imagery. From the MCD43B4 we extract a medium solar zenith angle per pixel, provided as a range 0-5, 5-10 degrees etc. Is it possible to extract the solar zenith angle that has been used in derivation of BRDF parameters? We want an absolute value per pixel, not a range.

We are using all cloud-free atmosphere-corrected surface reflectances available over a 16 day period to retrieve the best fit, best sampled BRDF model.  The input observations are at overpass times for Aqua and Terra.  But no, we don't pass on the exact overpass szns of each input observation that is used.  In V004 we provide NBARs (MCD43B4) at the median overpass time for each satellite for that period; in V005 we will only provide one NBAR at local solar noon for that location (which is what we give for albedo in V004 and V005) and it will be to the full degree -not binned into 5 degree bins (which is only done because we are always constrained by how many bits we can archive).  You can of course compute average szn for each satellite or local solar noon for each pixel off line, or you can use the MCD43B1 model parameters and provide NBAR at any explicit szn you'd like.  There is some sample forward code under tools on our web page to use to compute NBAR.  That is probably your best bet if you need view-angle corrected data at some specific time each day.

Images seem distorted at the edges of scans. Why, and what can be done?

The Bow-Tie effect is an artifact of the arrangement of sensors on the MODIS <http://www.sat.dundee.ac.uk/modis.html> instrument. This results from the fact that MODIS scans 10 lines at a time. The pixel size on the ground increases with distance from the satellite. This satellite-to-ground distance increases with scan angle, mainly due to earth curvature. This means that the pixels near the edge of an image are bigger than the ones in the middle. The satellite is configured to move forward 10km in the time that it takes to perform one scan. This is so that pixels in the middle of the scan line match up next to each other. However, at the edges the pixels are bigger, up to 6 times wider and 4 times longer. This causes over-sampling, i.e., the same bit is imaged twice. As a result, the image seems distorted near the edges.

The bow tie is a purely geometric effect. We do not "correct" for it in the Level 1B data. However, some downstream products (Level 2 or higher) may provide a correction for the bow tie effect. The effect should really be removed after all scientific processing of the data has been completed, i.e., after atmospheric correction and product generation. This lets the user know the appropriate look angle and solar angle for each pixel. An acceptable solution is to use a reprojection. The reprojection takes the geolocation product and uses it to produce a image with a standard projection. The reprojected image is free from the bow tie artifacts, the MODIS quicklooks are an example of this. The problem with this is that you aren't able to calculate solar and zenith angles easily.

An example of thebowtie effect (near the right-hand edge of a 250m band)

Why do white spots occur on the MODIS L1B images?

http://w3k.gkss.de/kof/software/modis_tutorials/L1b/subfolder/FAQ/A_sci_spot.html

Responsible Civil Servant: 
Dr. Jack Xiong <Xiaoxiong.Xiong.1@gsfc.nasa.gov>

Email mcst-webm@ssaihq.com to report any problem with the MCST Web Pages.

Last Update:  July 3, 2008