Hi, I am also working on this topic! I was wondering whether you have the GEE code and could share it?

Thank you in advance!

Hi

For the first question:

1- Call your interested image collection

2- filterDate

3- filterBounds

4- .map(function) for getting EVI values (using .expression) and .clip(geometry)

5- get .median() values

For questions 2 and 3, GEE has provided lots of functions and image analysis algorithms but unfortunately there are not any prepared functions to be used for smoothing or extracting phenology parameters and users have to program them. You can use TIMESAT guideline report to program your interested function in GEE.

Thanks! Yes I am working on it!

Thank you. Actually, I can solve Queation 1and 2. But I still thinking how to solve question 3. Does any one have some specific suggestions(which functions can be used in Gee)? Thanks! Eduarda Silveira Alireza Taheri Dehkordi

Yuanhi Zhu. Fro question 3 there is no function available in GEE! I a trying to solve it but still working on it! For example, one idea to capture the start of the growing season (SOS) is to calculate the percentile 25th of the annual curve and get its DOY. When I have the code I can share!

Thank you, Eduarda. If you figure out, please let me know (My email and gee account: [email protected]). I will be also working on it. If any progress, We can talk. Now, the only way I can think of is trying to convert the images to vector to get each pixel value by reduceRegion function.

Thanks again.

Hey Eduarda,

Im also working on a phenology approach to distinguish different agricultural land use. I found a code on stack exchange how to fit a harmonic model for a given time series of spectral values or derived metrics such as EVI. The code also produces a raster image with the harmonic coefficients of the harmonic phenological characteristics of each land cover for each pixel. This can be used as a classification input as done in a recent study of Derwin et al. (2019).

https://gis.stackexchange.com/questions/271454/harmonic-modis-trend-in-google-earth-engine

Best regards, Marcel

Hi Eduarda,

I'm working on something similar, and I couldn't do it in GEE. What I did was compile the time series in GEE, then I downloaded the data to my samples (points or polygons) and processed it in R.

The 'greenbrown' package calculates all these metrics for seasonal time series.

With this data you can adjust a classification model with the algorithm you prefer. What I want do to do now is take this model and apply it to a larger area. Recently an R package that integrates it with GEE was released , but I haven't tested it yet:

https://rdrr.io/github/csaybar/rgee/man/rgee-package.html

The values you mentioned (phenometrics) are different from the sine and cosine components of the harmonic function. The latter you can easily calculate in GEE as Marcel Mohr said.

Please, let us know if you made some progress.

Kind regards,

Guilherme

The real challenge is how to solve the problem of multiple growing seasons.

1. VIs can be smoothed by Whittaker or HANTS on the GEE (note that not weighted SG, very low efficient on the GEE). Article A robust method for reconstructing global MODIS EVI time ser...

2. The threshold based phenology Extraction algorithm can be easily implemented, if the problem of multiple growing seasons solved.

3. Phenological metrics evaluation can be done by phenofit or rTIMESAT on the local computer.

- https://github.com/kongdd/phenofit

- https://github.com/kongdd/rTIMESAT

Hi all

Maybe this paper introduces something related to question 3.

Detection of Crop Seeding and Harvest through Analysis of Time-Series Sentinel-1 Interferometric SAR Data 

 https://www.mdpi.com/2072-4292/12/10/1551 

Jiali Shang, Jiangui Liu, Valentin Poncos, Xiaoyuan Geng, Budong Qian, Qihao Chen, Taifeng Dong, Dan Macdonald, Tim Martin, John Kovacs and Dan Walters 

Remote Sens. 2020, 12(10), 1551; doi:10.3390/rs12101551 

Eduarda Silveira You may find answer from this attached paper.

Hi Silveira,

I am also working on your interest. But I have a solution for questions 1 and 2 but have no solution for questions 3. but I have a paper they completed it in GEE. I also contacted them but don't get any feedback.

The link of this paper is Article A dataset of 30 m annual vegetation phenology indicators (19...

If you have any solution before my comment, please let me know within my mail id - [email protected].

Thank you.

Regards, Arabinda.

for question 3, you can try to use this dataset in GEE

var phenologyData = ee.ImageCollection("MODIS/006/MCD12Q2");

I think that this interesting class by Nick Clinton on Time series analysis covers almost all your needs.

Slides

https://docs.google.com/presentation/d/1J1rUtf-bkfaJwYJY-tU17kzKI4U8FnF7Q2_VWqWdaak/edit#slide=id.g4954714e1_18

Video

https://www.youtube.com/watch?v=xQGNAYL7z80

transcript with code

https://docs.google.com/document/d/1mNIRB90jwLuASO1JYas1kuOXCLbOoy1Z4NlV1qIXM10/edit#heading=h.6f09b65xgrdt

Time-series NDVI/EVI data could be firstly extracted from GEE. Modeling of NDVI/EVI phenology could be done on MATLAB cftool (Curve Fitting Tool), where many model forms are provided.

Hello. Think the paper entitled 'Spatiotemporal image fusion in Google Earth Engine for annual estimates of land surface phenology in a heterogenous landscape' (Nietupski et al 2021), which contains GEE codes, may be adapted to the problem.

Hello. Maybe this work of a colleague is useful. Atm only published for MODIS, but exchanging the sensor is possible.

https://github.com/MBalthasar/Harmonics_and_z-score_GEE_code

Best

Sebastian

Hello, I did one for composite Landsat indices which may help you but you will need to extend to add phenometrics. Please see:

https://code.earthengine.google.com/739ab649097aea84052043cd6c63c905

if you want to reference this methodology, please cite:

Santos, F., Meneses, P., & Hostert, P. (2019). Monitoring long-term forest dynamics with scarce data: a multi-date classification implementation in the Ecuadorian Amazon. European Journal of Remote Sensing, 52(sup1). https://doi.org/10.1080/22797254.2018.1533793

The authors of this paper Article Mapping Phenological Functional Types (PhFT) in the Indian E...

Jesús N. Pinto-Ledezma Couldn't clone the codes from this paper. Supposing they removed public access.

code is not working for me

You can check our GEE community tutorial, It could be a useful starting point. https://developers.google.com/earth-engine/tutorials/community/histarfm-cloud-and-gap-free-landsat

Use the new band, i.e. the time band derived from the date information to determine when the Maximum Vegetation Index (EVI) (EVI_Max) occurs using the qualityMosaic('EVI') Composite method at a specified time period, for example 4 months for rice crops. Furthermore, when the SOS and EOS of rice plants can be determined with the assumption that EVI_Max occurs in the middle of the growing season (60 days after planting).