VIIRS Day- Night Band Cloud- free Composites March 3, 2015

Transcription

1 VIIRS Day- Night Band Cloud- free Composites March 3, 2015 Kimberly Baugh Earth ObservaIon Group (EOG) CIRES - University of Colorado, USA NOAA NaIonal Geophysical Data Center, USA Kim.baugh@noaa.gov Chris Elvidge - NOAA NaIonal Geophysical Data Center, USA Mikhail Zhizhin - CIRES - University of Colorado, USA Feng Chi Hsu - CIRES - University of Colorado, USA

2 Lights At Night Cities and human settlements Boats Industrial Sites Gas Flares Fires 2

3 NighVme Lights Composites The EOG Group at NGDC has a long history of making global annual nighvme lights composites using DMSP- OLS data. h"p://

4 NighVme Lights from Space Earth ObservaIon Group (EOG) at NGDC has been making nighvme lights products with DMSP- OLS since Prior to the SNPP launch in October 2011 the only global nighvme lights dataset was from the Defense Meteorological Satellite Program s OperaIonal Linescan System (DMSP- OLS) DMSP- OLS sensors have been flown on DMSP polar- orbiing pla`orms since the 1960s. Digital OLS data has been archived at NOAA s NaIonal Geophysical Data Center (NGDC) since 1992.

5 VIIRS Day- Night Band vs DMSP- OLS SpaIal ResoluIon The VIIRS DNB footprint is 45 Imes smaller than the nighvme DMSP- OLS pixel footprint! NighVme DMSP OLS 5 km footprint VIIRS Day / Night Band 742 m footprint

6 VIIRS Day- Night Band vs DMSP- OLS Quan%za%on: DNB is 14 bit versus 6 bit for OLS. Dynamic Range: Due to limited dynamic range, OLS data saturate on bright lights in operaional data collecions. Lower Detec*on Limits: DNB can detect dimmer lighing than OLS. Quan%ta%ve: DNB is calibrated, the OLS visible band has no in- flight calibraion. Mul%spectral: VIIRS has addiional spectral bands to discriminate combusion sources from lights and to characterize the opical thickness of clouds.

7 Comparison of DMSP- OLS with VIIRS DNB Attribute DMSP/OLS VIIRS/DNB Orbit Sun-synchronous, ~850 km Sun-synchronous, 827 km Nighttime Nodal Overpass Time ~1930 UTC ~0130 UTC Swath Width 3000 km 3000 km Spectral Response (full-width halfmaximum) Panchromatic nm Panchromatic nm Ground-projected Instantaneous Field of View Spatial Resolution (Ground Sample Distance) 5 km (nadir) / ~7 km (edge) ± km (Scan) ± km (track) 2.7 km; smooth data < km (Scan) < km (track) Minimum Detectable Signal W m -2 sr W m -2 sr -1 Noise Floor ~ W m -2 sr -1 ~ W m -2 sr -1 Radiometric Quantization 6 bit 14 bit Accompanying Spectral Bands 1 (10-12um) 11 at night / 21 at day Radiometric Calibration None On-Board Solar Diffuser Saturation In Urban Cores None 7

8 DMSP- OLS vs VIIRS Day- Night Band DMSP- OLS October 14, :30 VIIRS DNB October 15, :30 Note the lack of DNB saturaion in Bangkok. Also the increased spaial resoluion and lower detecion limits allow DNB to disinguish small roads and more isolated fishing boats.

9 DMSP- OLS vs VIIRS Day- Night Band OLS has saturaion in city centers and less spaial detail as compared to the DNB DMSP- OLS VIIRS DNB

10 DMSP- OLS vs VIIRS Day- Night Band A contrast stretch applied to the same VIIRS DNB image brings out the dimmer lighing features that aren t seen with the DMSP- OLS. DMSP- OLS VIIRS DNB

11 NighVme Lights Composites What are they? A nighvme lights composite is made to serve as a baseline of persistent light sources. Composites are made as an average of the highest quality nighvme lights imagery over desired Ime period usually monthly or annually. EOG group has made 4 temporal prototypes of nighvme lights composites using VIIRS DNB: April 2012, October 2012, January 2013, and May EOG group is producing current monthly DNB nighvme lights composites and will expand to remove ephemeral light sources in 2015.

12 NighVme Lights Composites What goes in? Only the highest quality nighvme data gets averaged into a composite Currently this is defined as DNB data that is: Cloud- free (using the VIIRS cloud- mask IICMO product) NighVme with solar zenith angles greater than 101 Not affected by moonlight (lunar illuminance < lux) Outside the region impacted by stray light (nadir solar zenith angles > 118.5) Middle of swath (DNB has increased noise at edge of scan) Free of lights from lightning Free of lights from South AtlanIc Anomaly

13 Using VIIRS DNB for NighVme Lights Composites Some DMSP- OLS algorithms could be reused Day/night/twilight flagging Zero lunar illuminance flagging Stray light region flagging Cloud algorithm (used M15 in place of OLS thermal band) Some algorithms needed makeovers Lightning detector (to work on 16- line scan) Terrain correcion for geolocaion New algorithms PoinIng correcion for geolocaion SAA filter Blurry lights filter (to remove reliance on cloud mask) Fire removal (taking advantage of other VIIRS spectral bands) Stray light correcion (port Northrup Grumman algorithm)

14 VIIRS DNB Composites First Asempt First prototypes made in Dec 2012 for low- moon nights in April and Oct Average radiance values were constructed on a 15 arc- second grid for data determined to be: Cloud- free Zero lunar- illuminance Out of stray light region Center aggregaion zones

15 Processing VIIRS DNB Aggregates: Flag Bands DNB DNB FLAG Red: DayIme Green: Terminator zone Blue: NighVme Aggregate SVDNB_npp_d _t _e _b05050_c _noaa_ops.h5 For each nighvme VIIRS DNB aggregate, a companion flag band is generated with bit- codes designaing: DayIme (pixel solar zenith angle < 96) Day- Night terminator zone (96 < pixel solar zenith < 101) Zero lunar Illuminance (< lux) Stray light region (90 < nadir solar zenith < 118.5) Lunar illuminance is a funcion of lunar phase, azimuth, and elevaion, which are also based on lat, lon, and Ime of each DNB pixel. This enire aggregate was flagged as having zero lunar illuminance and being affected by stray light.

16 Processing VIIRS DNB Aggregates: Flag Bands Aggregate SVDNB_npp_d _t _e _b05050_c _noaa_ops.h5 DNB DNB FLAG Blue: Stray light region Orange: Edge of Scan (DNB aggregaion zones 29-32). Black: This area is considered high quality nighvme data. Increased noise at edge of scan For each nighvme VIIRS DNB aggregate, edge pixels are discarded due to increased noise (DNB aggregaion zones 29-32). This enire aggregate was flagged as having zero lunar illuminance.

17 Processing VIIRS DNB Aggregates: Cloud Mask Clouds obscure the nighvme lights data. To avoid using data that has been impacted by cloud- cover, a cloud mask is needed. Cloud masks are generated using an NGDC algorithm comparing the VIIRS M15 band (10.763um) with modeled surface temperature. M15 Brightness Temperature DNB Note: Black areas at edges of scans in M15 data are no- data values due to onboard bow- Ie deleions. They do not affect the reprojected image. M15 Cloud Mask Clouds are impacing nighvme lights of Chicago by reducing the intensity of the lights and by blurring the spaial detail of the light features. Aggregate npp_d _t _e _b05050

18 Processing VIIRS DNB Aggregates: ReprojecIon DNB and flag data is reprojected into 15- arc second grids using terrain- corrected posiion informaion VIIRS M- bands and DNB have different sensor and aggregaion characterisics which require them to be reprojected independently. The cloud mask generated from M15 is merged with the flags generated from the DNB auer reprojecion. Flag from DNB Combined VIIRS flag Flag from M15 (cloud mask) Aggregate SVDNB_npp_d _t _e _b05050_c _noaa_ops.h5

19 DNB Average Processing: ComposiIng The 15- arc second DNB grids are masked to areas of highest- quality nighvme data using the flag band. Highest quality for this first asempt meant: cloud- free, zero lunar illuminance, dark nighsime, outside stray- light region, and middle of swath (aggregaion zones 1-28). The masked grids are then composited, creaing a suite of files including an average DNB image, number of cloud- free observaions used, and standard deviaion of DNB for each grid cell. Average DNB N Cloud- free ObservaIons

20 VIIRS DNB Composites First Asempt Composites weren t as sharp as expected. We suspected either the cloud algorithm and/or errors in geolocaion. InvesIgaIon revealed a known DNB poining error. NGDC received a table of esimated poining errors from L. Liao at Northrup Grumman, which were then matched with GEO LUT filenames recorded as an asribute in the DNB h5 files. Adding poining error adjustment to terrain correcion souware made huge improvement in composite feature sharpness.

21 VIIRS DNB Composite (Oct 2012) Before PoinIng Error CorrecIon Close- up of Los Angeles Basin. Toggle with next slide. NoIce westward shiu and increased spread of lighing features due to poining error.

22 VIIRS DNB Composite (Jan 2013) Auer PoinIng Error CorrecIon Close- up of Los Angeles Basin. Toggle with prev slide. NoIce westward shiu and increased spread of lighing features due to poining error.

23 VIIRS DNB Composites Second Asempt Second prototype made in mid for all low- moon nights in Jan It was decided to try using the VIIRS Cloud Mask for the next asempt to see if addiional blurriness was reduced. Composite sill wasn t as sharp as expected in some regions of the world. Some clouds seem to be evading the cloud mask resuling in blurry lights. Calgary, Canada. Jan 2013 DNB Composite.

24 VIIRS DNB Composites Current Asempt Processing May December 2014 data. AddiIonal algorithms being run are: Sharpness Index (remove blurry lights without reliance on cloud mask) Lightning filter South AtlanIc Anomaly (SAA) filter

25 Sharpness Index On the leu is a DNB image showing areas with blur induced by clouds. On the right is the sharpness index image. Blurry areas are dark and sharp lights are bright. By applying a threshold on this index it will be possible to screen blurry areas from the composite.

26 Lightning Filter Example of lightning streaks detected by the DNB. The streaks are sixteen lines wide, arising from individual scans. Removing reliance on a cloud- mask by using the blur index will make filtering for lightning signatures necessary for a clean DNB composite. 26

27 South AtlanIc Anomaly (SAA) Filter Example of high values in DNB due to high energy paricles in South AtlanIc Anomaly region. Red pixels were determined to be SAA hits because they exceed the average of neighboring pixels by more than 99%. March 31, 2012 off coast of Brazil 27

28 NGDC DNB Data Availability AcIve monthly product generaion started 5/1/14. The products are available at: hsp://ngdc.noaa.gov/eog/viirs/download_monthly.html NGDC also generates nightly mosaics in png and Google Earth Super- overlay formats hsp://ngdc.noaa.gov/eog/viirs/download_ut_mos.html 28

29 Nightly Global Mosaics Cover Single UT September 11, 2013 Sunlit Missing 29

30 Next Steps for VIIRS DNB Composites Filter out lights from aurora Normalize historical data from region impacted by stray- light so its usable in the composites use NG algorithm? Implement the sharpness index filter to remove reliance on VIIRS cloud- detecion. Filter out ephemeral lights from boats and fires to create a VIIRS DNB Stable Lights product Nile Delta

31 SeparaIng Fires from Lights in DNB SIll in algorithm development - R&D SeparaIng fires from lights using NGDC Nigh`ire product The image on the leu is the raw DNB. The image on the right shows the masking of biomass burning pixels from the Nigh`ire (VNF) data.

32 Stray light correcion algorithm from Northrup Grumman This algorithm was implemented at the IDPS in August 21, We will likely need to implement the algorithm at NGDC and apply it to archive data acquired prior to that.

33 DNB Atmospheric CorrecIon In development R&D The loss of signal in the DNB due to atmospheric absorpion and scaser is both substanial and highly variable, in the range of 15 to 60%. We are working on an atmospheric correcion for the DNB that uses MODTRAN to esimate the transmissivity of the atmosphere in the DNB. We parameterize MODTRAN using atmospheric profiles generated from ATMS data, which are collected simultaneous to the VIIRS. Specifically, we will use atmospheric pressure, temperature and relaive humidity profiles generated from ATMS data using the MIIRS processing package (NOAA, 2013). The MODTRAN runs are computaionally intensive, therefore the correcion will only be run on pixels that are entering the monthly composites. 33

34 QuesIons? Contact: (USA)