Satellite Image Analysis - Science topic

Yusuf Mamman , Andongma Wanduku Tende Yunis Boga

Relative radiometric normalization using pseudo invariant features (PIF) in Google Earth Engine can be done by following these steps:

1. Add the input imagery to Earth Engine.

2. Identify potential PIFs by exploring the image and selecting features that do not change significantly in the image over the time period of interest.

3. Label PIFs and assign each one a unique identifier.

4. Select the sample points to be used for radiometric normalization.

5. Calculate the mean and standard deviation of the radiometric values for the sample points.

6. Calculate the radiometric value of the PIFs in the input imagery.

7. Calculate the normalization factor for each PIF by dividing the radiometric value of the PIF by the mean and standard deviation of the sample points.

8. Apply the normalization factor to the input imagery in order to achieve relative radiometric normalization using the PIFs.

Script:

// Define a function to calculate the relative radiometric normalization using PIF

var relativeRadiometricNormalization = function(image){

// Calculate the normalization factor

var normalizationFactor = image.normalizedDifference(['B7','B5']).select('B7');

// Select the bands to be used for the normalization

var bands = ['B2','B3','B4','B5','B6','B7'];

// Map the normalization over the bands

var normalized = image.select(bands).multiply(normalizationFactor).divide(100);

// Return the normalized image

return normalized;

};

// Apply the function to the image

var normalizedImage = relativeRadiometricNormalization(image);

Hope this will work.

Cara Rosa,

There are no universal thresholds to classify urban heat islands because each city will experience its own specific climate and environmental constraints. Since you are dealing with a single city, I suggest you start by determining the lowest and the highest temperatures in your data set, and then divide that range (Tmax - Tmin) in 3 equal intervals. For instance, if Tmin = 18 and Tmax = 24, your intervals would be 18 to 20, 20 to 22 and 22 to 24 C. In that case, all areas falling in the lower interval could be labeled 'cool', the areas falling in the next class could be called 'warm', and those belonging to the higher interval could be indicative of 'hot' conditions.

You should make sure that your data set covers a region around the city large enough to include agricultural fields or forests: those areas would provide you with a baseline environmental temperature away from the urban center. An obvious extension of this approach would be to map the city areas in single degree intervals, from blues through to greens, yellows, oranges and red. That will clearly indicate which areas are hotter.

A better approach would be to start from a preliminary question: why do you need to make a map of the urban heat island effect of Naples in the first place? If you were concerned by the health effect of temperature on morbidity and mortality of the inhabitants, for instance, then your temperature thresholds should be driven by medical rather than purely statistical considerations. Similarly, if your underlying concern were energy expenditures (cooling during the summer or heating during the winter), then your thresholds should relate to the corresponding critical rates of energy expenditure. In other words, your approach should depend on your ultimate goal.

Remember also that

- the urban heat island is quite time-dependent: it varies with cloudiness and synoptic conditions such as sea-breeze on a daily time scale, it is much more noticeable during the winter than the summer, and it may evolve on longer time scales, depending on the rates of urbanization and industrialization;

- land surface temperature is quite dependent on altitude, so you might want to acquire a topographic map of your area and look at the correlation between these two parameters;

- a land use map of the region will be useful to properly interpret your results, as the hotter area may not be the city center but some industrial area, depending on their relative rates of energy consumption.

I hope these comments may help you in your work. Michel.

The density clustering with wavelength clustering algorithms and Clustering by Wavelet Analysis may help your work

if you have ArcMap 10.8 then you can create easily, you can use toolbar> production cartography> and click Grid and graticules designer.

In the Landsat 8 Surface Reflectance dataset that is given on Google Earth Engine, do I have to do any additional preprocessing as it is already atmospherically corrected or do i have to just scale it using the scaling factor?

If you use Tier 1 Surface Reflectance, 'LANDSAT/LC08/C01/T1_SR', mentioned by Amirmohamad-Abasi, you don't need to do any preprocessing . You can also use Tier 1 Level 2 product "LANDSAT/LC08/C02/T1_L2" to get surface reflectance data.

What is the difference between the Landsat Surface Reflectance datasets and Landsat TOA Reflectance datasets?

TOA Reflectance datasets need atmospheric correction.

You may be interested in:

In my study area I have 153 stations. I have monthly precipitation data for all these for 40 years. I should I find SPI for WHOLE area? I know R studio.

Hamed Heydari Jose Luis Ornelas-de Anda Stavros Kolios Yuke Zhou

Dear Pritiranjan Das

Here are some of the most prominent models for simulating urban growth. Besides, Jiří Kroc has provided a very good approach as well.

Thanks and regards

Gowhar Meraj

Hi!,

Given that the crop seasons are known and the availability of sentinel data, maybe instead of post-processing you can perform a pre-processing, select a date or dates when the crops are still not developed and classify the forests, then create a mask of forest areas and apply that mask to an image or images for the crops classification. Depending on the complexity and types of crops you may have to anlyse more than one date but in this way you remove the effect of forests.

Hope it helps, I have solved this way a similar problem I had with urban and soil with stubble mixed signatures.

Regards,

Soledad

I used MODIS LST day and night temperature. I suggest you, you can use this.

Surface temperature can be calculated using the two thermal bands of the Landsat 8 , this can be calculated by applied the radiometric calibration via ENVI software.

Most GIS and RS software have the standard classification methods (ArcGIS, QGIS, ENVI, Eedas etc).

As a prerequisite dimensions of all rasters must be identical i.e. rows and columns for MCE.

TerrSet facilitates image resampling. A simple approach can be to use 'Window' tool available therein for the purpose. Just provide reference image to serve as desired spatial extent and clip all the input rasters accordingly.

Eneche Patrick Samson Udama, glad to know this discussion page helped. Appreciate your comments.

I was recently attending some NASA Applied Remote Sensing Training (ARSET) on Land Surface Temperature and Urban Heat Island (UHI) and Night Time Satellite Imagery Applications. If someone is interested they can follow the links:

Hi Arka,

If you have the coordinates of the area which you need, you may use the following code to extract the images combined together for long/short time periods.

However, you have to make sure about the size of the image/ area as sometimes, it may give error from the server-side in GEE, if there are a huge number of images in one go. In that case, you may try to take it in small sets.

Hope this helps you.

Cheers

More Precisely,

Let's, Sensor, (Sw x Sh) = 13.2 x 8.8 mm

Focal length, F = 8.8 mm (rear focal length, not 35 mm equivalent)

The flight height, H = 100 m

Video resolution, (imW x imH) = 4096 x 2160 (in pixels)

Then,

Ground Sampling Distance, (GSD) = (H * Sw) / (F * imW) (in cm)

Also, Image width footprint on the ground = (H * Sw) / (F ) (in meters)

GSD Calculator: https://www.3dflow.net/ground-sampling-distance-calculator/

Cartosat-3 is an advanced Indian Earth Observation satellite built and developed by ISRO, which will replace the IRS series. It has a panchromatic resolution of 0.25 metres making it the imaging satellite with highest resolution in the world and Mx of 1 metre with a high quality resolution which is a major improvement from the previous payloads in the Cartosat series. Potential uses include weather mapping, cartography or army defence, and strategic applications.

For automatic lineament extraction, you can achieve this by applying a combination of the following softwares: Arcmap 10.0+, ENVI 5.1, PCI GEOMATICA 2016, ROCKWORKS 16.

STEP1: Use ENVI software to perform principal component analysis (PCA) on the spectral image band of Landsat8.

STEP2: GEOMATICA is used to perform an automatic lineament delineation from the PC1.tiff generated from ENVI. This is done with the aid of lineament extraction algorithm in the Tool from toolbar

STEP3: ARCMAP is used for handling extracted lineament (splitting compound line into simple lines, editing lineament attribute, exporting lineament as cad file

STEP4: ROCKWORKS is used to process the exported lineament Cad file to determine the trend of the lineament and hence generating rose diagram.

For automatic lineament extraction, you can achieve this by applying a combination of the following softwares: Arcmap 10.0+, ENVI 5.1, PCI GEOMATICA 2016, ROCKWORKS 16.

STEP1: Use ENVI software to perform principal component analysis (PCA) on the spectral image.

STEP2: GEOMATICA is used to perform an automatic lineament delineation from the PC1.tiff generated from ENVI. This is done with the aid of lineament extraction algorithm in the Tool from toolbar

STEP3: ARCMAP is used for handling extracted lineament (splitting compound line into simple lines, editing lineament attribute, exporting lineament as cad file

STEP4: ROCKWORKS is used to process the exported lineament Cad file to determine the trend of the lineament and hence generating rose diagram.

Does the Wavelet transformation has been used at any stage of your research?

yes, first you have to check your comailed database.

Dear Leônidas Luiz Volcato Descovi Filho ,

Yes, I agree with you but it is not always possible to get reference geodetic bases of the first order.

Hi Mannan,

Are the qanat's track superficial or underground? if they are superficial, you can use thermal bands. If the qanat are underground, I think you can use Ground Penetrating Radar for this purpose.

Good Luck,

Reda

For RS and GIS applications such as yours, ERDAS IMAGINE, ENVI, ARCMAP, IDRISI may be of great help. They are dedicated remote sensing and GIS tools that provide platforms for a great variety of applications such as mapping, and change analysis. Since your study is mapping roofing materials, all of the above mentioned softwares provide good platforms for your work. There are thousands of videos available online for learning and understanding supervised classification. Data preparation is equally important as well. Refer "Introductory digital image processing" by Jenson for more detailed information on both classification and data preparation.

and also in c++

Dear D. Object oriented segmentation with sub-meter resolution satellite imagery and using SW like eCognition sounds very promising! Good Luck!

dear try these links

If you are willing to pay the costs for high resolution imagery like Worldview or Ikonos repeatedly, it certainly is possible at an excruciating amount. With opensource image its really difficult because of the coarse resolution. You can try using Sentinel-2 which is 10m to may be monitor the drainage systems. But the problem is in the rural areas these are very narrow and would not even comprise a single pixel on certain occasions. The best would be to use UAVs for monitoring, it wouldn't require a very sophisticated sensor so you could manage your budget also.

Dear

there are a large number of softwares can do that such as envi, arc map

regards

you can use Envi software and then u can use unsupervised classification

Hello!

Yes! the careful examination/interpretation of remote sensing images by using various false color composites will help us to collect the signature of land use/land cover classes effectively.

You may refer following article for information about, how to achieve better classification accuracy.

All the best

You need to know the seasonal transition parameters in order to extract phenology data out of remote sensing time-series data. I recommend you yo check TIMESAT (http://web.nateko.lu.se/timesat/timesat.asp) for the extraction of these parameters which are mainly: beginning of season, end of season,length of season, time of middle of season, maximum value, amplitude etc. You can model the same approach or try different fittings for your time series data with R but the main idea is explained briefly in the software manual.

After extracting phenological data you can build your own model between yield and vegetation indices.

You can also check our study, performed with Landsat data, which is not totally aiming yield prediction but stating that the use of phenological data for agricultural applications helps to improve accuracy:

https://docs.opencv.org/3.4/d9/d51/classcv_1_1ximgproc_1_1DisparityWLSFilter.html

I think you are better to visit this website .I hope it is useful for you

You can also trip MODIS or ASTER products.

Check the links below;

Wishes

Dear @Muhammad Budi

You need to work with SLC not GRD when you study about displacement and deformation because only SLC have phase and amplitude while GRD have only amplitude.

My Greeting

You may find the link below:

Julius, I still being able to access https://viewer.nationalmap.gov/tools/rasterconversion/Convert_JP2_Data_to_GeoTIFF_with_GDAL.html

You could write a batch (.bat) file with something like:

for %%N in (path/to/your/sentinel2/*.jp2) DO C:/Program Files/QGIS 2.14.5/bin/gdal_translate.exe -of GTiff C:/temp/sentinel2/%%~nN.tif

@ECHO OFF

SETLOCAL EnableDelayedExpansion

SET mypath=C:\temp\sentinel2\

FOR /F %%i IN ('DIR /B %mypath%*.jp2') DO (

SET infile=%%i

SET outfile=!infile:.jp2=.tif!

C:/Program Files/QGIS 2.14.5/bin/gdal_translate.exe -of "GTiff" %mypath%!infile! %mypath%!outfile!

)

You'll need to adjust it and redirect to your gdal_translate.exe and jp2 folders.

Yes. You can find our public available dataset EVLabSS dataset, which is designed for semantic segmentation of high-resolution images. Our website is: http://earthvisionlab.whu.edu.cn/zm/SemanticSegmentation/index.html

Hope it help your research work.

Normalization in the context of remote sensing refers to a process to make something comparable with something else.

For example, if you are doing a change detection study, you may be obviously looking at two or multiple Landsat images of a particular location. So how can you make sure that the changes that you are observing in these different landsat images are caused by the land surfaces or are the affects of atmosphere? Therefore, in order to compare two images from two different time periods for a land cover change detection study, you would first remove the effects of atmosphere by 'atmospheric correction'. This is a kind of normalization. To give you another example, if you are analyzing two different satellite images with different spatial resolutions, you would employ some kind of pixel resampling technique in order to make the two images comparable. Yet another example, lets say, two satellite images were acquired with different sun elevation and azimuth. In this case, you may try out some sun angle corrections, before comparing the two images. This is also a kind of normalization. There are several other normalizations but the aim is the same.

Q. How can you make sure that the target (e.g., vegetation) which you are studying has indeed changed over time or it looks changed only because of the effect of atmosphere or sun/sensor geometry or something else?

Ans. Normalization!

Joan, just wanted to share an update. All the Landsat 8 night time scenes that were acquired in the past are back online available for download on the Earthexplorer database. Out of the 17607 night time scenes acquired with Landsat 8 globally, there are around 5 scenes covering Barcelona for the time period July-Sept. 2015.

Cheers!

Both Envi and Photomod could also help u for this work:

Dear Venkata,

Your question is too vague to be able to assemble a meaningful answer. Please provide more details on your problem and in particular address the following points:

- What is the origin of your stereo images (ground-based, airborne or space-borne) and what is their spatial resolution?

- What is the nature of your target, and is it flat or rough (i.e., if it is a land surface, does it include topography)?

- Have these images been co-registered so that pixel (i, j) in both images correspond to the same point on the target?

- What are the format and mathematical nature of your stereo images (Black and White, RGB or multi-spectral; binary, integer, floating point)?

- What method did you use to generate the 0/1/2 values included in your 'similarity.xlsx' file? Specifically, the sum of squared differences between the values retrieved from two images would yield a single number, so I don't understand how you could map that...

- Also, what is the intended meaning of the scores 0, 1, and 2?

Once your problem and your approach are better understood, it may be possible to provide some guidance about how to proceed. Cheers, Michel.

Crop estimation or monitoring is done best with high-resolution polarimetric SAR data. It allows you to decompose the backscattered signal into different scattering mechanisms (see reference 1). PolSARpro is a great open software for that, but currently, there are no free full-polarimetric scenes available. Sensors which support it are TerraSAR-X, Radarsat-2 and ALOS PALSAR-2. For the first two, you can apply for academic access to a limited number of scenes (see references 2 and 3).

Generally speaking, there is no universal workflow which can be applied. But having field samples of crop types and volume estimates would surely be a good point to start with. Then, findinig some way to link the backscatter to the expected information is the next step.

There is a great tutorial by SAR EDU on crop type classification (see reference 4). You can access it after free registration.

If you rely on freely available data, Sentinel-1 would be the best alternative due to its high repeat frequency in many regions and at least dual-polarization (VV/VH). Unfortunately, c-band is not optimal for vegetation and crop types, but if your fields are large (spatial resolution is 10m) you can make time-series analyses on the backscatter of different plots and correlate it to your field samples in order to estimate the crop types.

Besides that, I can recommend going through some review papers on the topic which often show the pros and cons and present 0possible applications of all kinds of sensors (see references 5-6). 

Hassan, there are appropriate missions and sensors to study AOD like Aqua, Aura, CALIPSO, MISR, MODIS, NPP, Terra, VIIRS to name a few. I agree with Rangarajan, Landsat is not the best for AOD studies. I would strongly recommend you to follow some basic training on application of remote sensing in air quality studies. I have attached some links which you will find helpful. 

I have noticed that USGS has started reporting 'Combined RMSE of the geometric residuals (meters) in both across-track and along-track directions measured on the GCPs used in geometric precision correction' in the metadata (MTL.txt) file. This is very useful information. As you will read in the attached link, the geometric accuracy is influenced by the kind of Level-1 (L1-TP, L1-GT, L1-GS) landsat data that you are working with, which in turn is affected by 1) clouds 2) GCPs 3) ephemeris info. Another useful link is the 'GCP search link' provided by USGS. You can use the 'GCP search link' to extract ground control point for your area of interest and geo-reference Landsat Level-1 data. Links are attached.

You can use the Semi-Automatic Classification Plugin for QGIS. It allows user to download, pre-process and work with Landsat, Aster and Sentinel data.

It has a very good documentation and a number of tutorial are available at:  https://fromgistors.blogspot.com/p/semi-automatic-classification-plugin.html

Moreover, a few month ago, NASA had a webinar showing how to use this plugin: https://arset.gsfc.nasa.gov/land/webinars/advanced-land-classification

Finally, as it is developed in qgis, you can easely use it toghether with other algorithms, such as the already mensioned i.topo.corr or i.smap by grassgis.

Hi 

Obviously you have data from two different sources, then you have to adjust these data for altitude in order to obtained the adjusted combined data for gravity of that region.  

Hi Stavros,

Obviously the space view perspective grid is natural choice for products derived from geostationary platforms and thus standard in EUMETSAT. I have the same problem with the H05 product of H-SAF. I use very frequently cdo, but in this case you can not do very much with it. I wrote to the H-SAF with this question and I received the attached file.

 Best wishes, Hr.

Dear Akash,

If you need reliable quantitative information on the aerosol optical thickness (AOT), look at the atmospheric products generated by the MISR Science Team at NASA JPL. MISR is a multi-angle, multi-spectral instrument on-board the Terra platform. It has been operational since February 2000 and is still generating data today. The main purpose of that mission is precisely to characterize clouds and aerosols.

For more information on the MISR instrument itself, and for an exhaustive list of publications using MISR, visit

If you want to download data products, you can get them for free from the NASA Atmospheric Science Data Center (ASDC), which is accessible here:

Good luck in your investigations. Michel.

I think Abdelaziz gave good advice. The data in the L8 image are scaled to 16b integer to keep the size of the files down.

Reacaling coefficients are in the metadata files.

Info on how to do that here:

Nice Firoz ji...

Mohamed, Very thanks MODIS was a great suggestion!

As many others have stated, the blue wavelength penetrates the water the best which is why Landsat 8  band 1 (ultraviolet/blue) is called the coastal/aerosol band.  This band is affected most by atmospheric scatter (Rayleight scatter) so it also has a great deal of atmospheric 'noise'. 

Hi Mohan,

(1) You can use band math in ENVI to eliminate the negative values:

-This is the post-FLAASH  equation:

(B1 le 0)*0+(B1 ge 10000)*1+(B1 gt 0 and B1 lt 10000)*float (b1)/10000

- The post-DOS equation is similar but use 1 instead of 10.000 like this

(B1 le 0)*0+(B1 ge 1)*1+(B1 gt 0 and B1 lt 1)*float (b1)/1

(2) You need to collect spectra profiles in-situ at the same time that the pass of Landsat to make comparisons...

good luck!

Santiago 

Hi Vikas, thanks for your answer. Could you please give me some suggestions on books/papers to refer for spatio-temporal filtering of SAR images?

Thanks!

I would recommend using the panchromatic band (i.e. one single channel that covers almost the entire visible spectrum) in the process of pansharpening solely for image interpretation and vizualization purposes. As PS will modify the radiometric information contained in the bands, you should not use these in quantitative spectral analyses (e.g. calculation of a vegetation index) thereafter.

You may update SNAP and try again. 

You can get Sentinel-1 data from scihub.esa. Requires only registration (And most likely, non-commercial use). As Sentinel-1 has just become operational the archive is not very extensive but should grow quite quickly.

You can set request data-access propospal on Alaska Satellite Facility. Some data open access. For ALOS-PALSAR you must be a resident of the United States to receive approval for data access.

You can get data from UNAVCO SAR data archive.

And you can get data from earth.esa.int. To access the data, you must provide a project. To preview scenes, use EOLi esa. ENVISAT ASAR the best option.

Although tree species level identification has always been challenging if your are working at 10 m spatial resolution.

1. So Before going for individual tree species level analysis using Sentinel-2. You must go for some preliminary analysis using Sentinel-2 data. Sentinel -2 provides 10 m spatial resolution for RGB and NIR bands while it provides 20 m spatial resolution for band 5, 6, 7 and 8A in Vegetation RedEdge category. 

2. Since RedEdge bands has capability to discriminate different tree species based upon their spectral variability. Therefore, Sentinel-2 can be a powerful opensource platform for these kind of analysis. 

3. If you do re-sampling of RedEdge bands to 10 m and create the Layer Stack of Red, Green, Blue, NIR and RedEdge Bands and then you can check the spectral signature of different trees species (it could be some homogeneous regions) for different band combinations. 

4. For this you must have some ground data. If you have a then you can cross check these spectral signatures and then you can go ahead for either object based or pixel based supervised classification procedures. 

 Hi Ran,

While ordering WV-3, did u unselect the DRA (Dynamic Range Adjustment) option?

If u received the data with DRA enhancement applied on it, then pre-processing cannot be performed. Even if radiometric correction is performed, it will not produce correct values

So, my suggestion is kindly check this.

Please check my communication with Longbotham, DigitalGlobe.

Dear Daniel,

Firoz is right: all (or most) trees look 'green' from space when the leaves are out, and it is not clear that their spectral response in the near-infrared is much different from species to species. You may find a somewhat higher discriminating power in the mid-infrared, where there may be more subtle variations due to the presence of components in variable concentrations such as lignin and cellulose, which exhibit specific vibrational spectroscopic signatures. Of course, this may not be applicable to Sentinel-2: that instrument is great to investigate environmental issues at fairly high spatial resolution, but it is not a hyperspectral sensor.

Instruments like WorldView-3 and others provide a better spatial resolution than Sentinel-2, but feature only a limited number of bands too. It may be worth exploring to what extent these instruments may be useful for your purpose, but don't expect miracles: you may very well reach the conclusion that tree species can't be distinguished this way.

On the other hand, if you want to pursue hyperspectral investigations, consider these leads:

Databases of spectroscopic signatures do exist, but they are often held by projects rather than widely available. One worth mentioning is the 1985 NASA study by Bowker et al., which is available here:

Check also the NASA Jet Propulsion Laboratory site concerning the airborne AVIRIS instrument:

Dr. Robert O. Green, the Principal Investigator for AVIRIS (http://scienceandtechnology.jpl.nasa.gov/people/r_green) and his colleagues have accumulated lots of hyperspectral measurements and generated a number of publications over the years.

Another researcher you may want to follow is Prof. Susan Ustin (http://lawr.ucdavis.edu/people/faculty/ustin-susan), as she has also worked extensively with hyperspectral data for ecological applications.

Other projects that may hold spectral signatures of different vegetation types (including trees) might be the LTER or FluxNet sites, for instance.

Check also the data generated by the Hyperion mission, which acquired over 200 spectral bands from space. In that case, you will need to assess to what extent the small signatures that may exist at the surface may still be distinguishable after the solar radiation has been transmitted through the atmosphere, before reaching the instrument:

Lastly, if you want to (or have to) use Sentinel-2 data, then you may be restricted to differentiate between evergreen and deciduous groups of species (using time series of observations), for example. Good luck, Michel.

keep accuracy  90 to 95%

Dear Dr. Firoz,

Please check the attached file which is an article about how to extract bathymetric map from Landsat 8.

Dear Sumit,

As Artem indicated earlier, the answer to your question depends mostly on the ultimate purpose of your investigation, as well as on the required accuracy of the information generated. From a purely logical point of view, however, it makes much more sense to first mask the clouds and cloud shadows at the high spatial resolution (30 m) of the Landsat data, than after aggregation into a low resolution product. First of all, it is much easier to detect and classify clouds and cloud shadows at high resolution. Second, the area contaminated by clouds and cloud shadows in the low resolution product will be larger than in the original Landsat data.

It is not clear why you are concerned by completing this task 'fast': the execution time largely depends on the performance of the hardware available, and somewhat on the software efficiency. In any case, new faster hardware will become available sooner or later. Except for a small number of time-critical applications, you should generally be much more concerned by the quality of your outputs instead. Another way to state this is that 'fast' is generally incompatible with 'safe' (or 'good quality').

Lastly, remember that there are various satellite-based instruments of Earth Observation that generate data at spatial resolutions around 250 m, including MODIS, MISR, MERIS, or Sentinel-3, and that those instruments provide a much more frequent coverage of your area of interest than Landsat. In other words, you may want to evaluate whether downgrading the spatial resolution of Landsat data by a factor 69 or so is really the best approach.

Cheers, Michel.

see this attachment as pdf file CMUSERSGUIDE.pdf

I am agree with Firoz Ahmad ...good luck

Great discussion. We have found the use of drones to often add unnecessary complexity and make the analytics challenging to impossible. While drones may be useful to you,  as detailed above in other answers, weigh your options carefully and align your research needs and applications with the available technologies. We have found often renting a pilot and airplane for 1 hour can produce far more reliable high resolution nadir or oblique images even with handheld Nikon D700-D812 camera's shot out of windows of small airplanes. Even mosiaced and georeferenced  oblique imagery when compared to drone imagery can produce significantly better imagery, more useful and comparable to high resolution airborne multispectral magery from stable airborne plateforms. Drone imagery has been challenging to mosaic, scale, georeference, control for platform instability, color normalize from adjacent flight lines, and for these reasons has not been particularly useful for measurements. We've opted for small aircraft very high resolution multispectral imaging, and are moving to very high resolution, full spectrum hypersphaetrael imagery complimented with Nikon 36 mp oblique imagery.

For SPOT 5 images, you could use the Airbus DS GeoStore directly:

But Airbus already has a few partners within Malaysia, maybe you find a way to cooperate with any of them for your needs:

Depending on your exact need for resolution and repeat cycles, you could also give the freely available Sentinel-2 or Landsat-8 data a chance to fulfill your needs:

Aloha all,

Please look at these documents:

Ocean Color - MODIS-Terra Ocean Color Reprocessing 2014.0 - Global mean deep-water

Compare multiple sensors between coastal and open ocean:

Comparative Analysis of Chlorophyll-a Distribution from SeaWiFS, MODIS-Aqua, MODIS-Terra and MERIS in the Arabian Sea

Hope this help,

Mahalo,

Roberto