Thanks for sharing, I would like to know if you have already tried using the NICFI, Landsat-8 and Landsat-9 planet images and the WRI ground cover instead of the ESA LandCover you used??
Excellent work presented. My question is whether these same results can be extrapolated to forage crops for animals. I would like to estimate the biomass of that forage crop. Thank you very much. I'm from Formosa, Argentina
The main difference between GEDI L4a and L4b AGBD datasets is that L4a provides footprint-level predictions, while L4b offers gridded estimates of AGBD with greater accuracy. I will also prepare a tutorial with Level 4a in the near future.
In case of large areas, "Output of image computation is too large", necessarily I can reduce the number of bands, which can affect the accuracy or better reduce the size of the area?
Validation focus on the RF model based on the training data set (split for training and validation). However, proper accuracy assessment should use an independent dataset. In most cases, researchers report only the model validation. However, it is better to report both validation and accuracy assessment (e.g., in the case of land cover mapping).
@@couragekamusoko5689 Thank you for the answer. I have one more question that Could it be possible for us to determine the number of trees ? Which number is better?
@@vietttt0104 The optimal number of trees in the Random Forest depends on your data set. Generally, the more trees you use the better model results. However, the improvement decreases as the number of trees increases. This means that at a certain threshold point, the prediction performance will not improve. Therefore, it is better to tune models in order to find the best trees. Tuning model parameters is easy in R or python. Unfortunately, I have never done it in Earth Engine.
My sincere apologies for the late reply. This depends on your research. I encourage you to use explainable machine learning techniques in addition to RSME. This is because a machine learning model can achieve high accuracy by simply memorizing patterns in the data. Therefore, relying only on model performance evaluation and prediction measures is questionable.
Hey Courage! what a great tutorial! this is incredibly useful.
this is great work absolutely amazing
Thanks for sharing, I would like to know if you have already tried using the NICFI, Landsat-8 and Landsat-9 planet images and the WRI ground cover instead of the ESA LandCover you used??
Hi. Not yet. I will try to use these datasets in the next tutorials.
Best regards
Excellent work presented. My question is whether these same results can be extrapolated to forage crops for animals. I would like to estimate the biomass of that forage crop. Thank you very much. I'm from Formosa, Argentina
Nice work done.
Thanks for sharing
Can i ask why are you using GEDI level 4b product and not level 4a and what is the difference?
The main difference between GEDI L4a and L4b AGBD datasets is that L4a provides footprint-level predictions, while L4b offers gridded estimates of AGBD with greater accuracy. I will also prepare a tutorial with Level 4a in the near future.
great tutorial!!!
In case of large areas, "Output of image computation is too large", necessarily I can reduce the number of bands, which can affect the accuracy or better reduce the size of the area?
Puedes compartir el Scritp?
Hello. Thank you for the model. But, I am having some errors to my study area. Can you please help me?
thank you for your sharing, where can we find the source code please
I got confused about accuracy assessment and validation. What is the difference? do we need to do both? Thank you for this work
Validation focus on the RF model based on the training data set (split for training and validation). However, proper accuracy assessment should use an independent dataset. In most cases, researchers report only the model validation. However, it is better to report both validation and accuracy assessment (e.g., in the case of land cover mapping).
@@couragekamusoko5689 Thank you for the answer. I have one more question that Could it be possible for us to determine the number of trees ? Which number is better?
@@vietttt0104 The optimal number of trees in the Random Forest depends on your data set. Generally, the more trees you use the better model results. However, the improvement decreases as the number of trees increases. This means that at a certain threshold point, the prediction performance will not improve. Therefore, it is better to tune models in order to find the best trees. Tuning model parameters is easy in R or python. Unfortunately, I have never done it in Earth Engine.
Thank you very much. Could you please share or in other tutorial the tuning part (in Python)
@@vietttt0104 I do not have a tutorial in python yet. I only have in R
What the max number of RMSE can be tolerated so we get the ideal model prediction?
My sincere apologies for the late reply. This depends on your research. I encourage you to use explainable machine learning techniques in addition to RSME. This is because a machine learning model can achieve high accuracy by simply memorizing patterns in the data. Therefore, relying only on model performance evaluation and prediction measures is questionable.
@@couragekamusoko5689thank's for the explanation, i'll try your suggestion 🙏
Thanks for the tutorial that was very helpful.
Can you also please add a note on how to get cumulative biomass of the study area.
Hello Courage. Can we use GEDI L4B data for mapping AGBD for the year 2016? Thanks in advance. Can you please share your email?
GEDI satellite was launched on December 5, 2018; so there was no GEDI data in the year 2016