AI and computational education

AI language tool align with lower cognitive skills in the BLOOM's taxonomy. So, education can focus on higher level BLOOM taxonomy. 

Check BLOOM's taxonomy, verbs

Laplacian matrix in transformer, for feature extraction?

 Laplacian matrix in transformer, for feature extraction? 

P132H Mpro

 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8923085/pdf/41422_2022_Article_640.pdf

MITRE digital public good data

 synthetic patient data

https://synthetichealth.github.io/synthea/

https://www.mitre.org/news-insights/impact-story/mitre-created-synthea-designated-digital-public-good

Will Stuart thesis

Focus on Chattanooga region

Q1: tree canopy change

Q2: carbon sequestration and biomass

Q3: current canopy distribution

data, June and July only, Landstat 5 Thematic Mapper (TM)  and Landstat 8 operational Land image (OLI). Three data sets,  PlanetScope, Sentinel Imagery, Skysat imagery

Global carbon cycle: geological carbon cycle vs biological carbon cycle

biological carbon cycle is sentive to anthropgenic pressures. 

human impacts on temperaretue forests: example of Caolorado. 

urban forests: in addition to enviroment factors, urban forest add 'sense of place' as culture values, reduce volume of stormwater, filter air and reduce urban noises. Will claim that urban forest may ennace carbon sequestion rate at higher rate. Ref: Albireo 2020, Garvey 2022, Morreae. 

Trees on the edge grow faster than tree in the interior for temporte forest. (Qin, this may be explained due to growth competition pressure in the center versus the edge). So, urban forest are often at 'edge' forest. 

Q: Unsupervised: Isocluster was used. 

Red-edge: 

NVDI, concept of vegetation indices, using near-infrared (NIR) and red spectrum 

passive sensors, spatial, spectral and tempral resolutions. 

sentinel - 20m, planetscope - 3m, NAIP-60cm. 

Q: how are image strips determined? 

Landsat data was presented first. 

How training data is verified? SVM 

Q: unsupervised classification? what methods? 

for historal images, google-earth histore can be used to verify the results. 

Conclusion: 43% loss of urban forest in Chattanooga, impermiave surce increase by 134%. 

NAIP data from 2018

Digital surface model (DSM) generated from 3DEP LiDAR. 

Vegetation indices: NDVI, GNDVI, SAVI, RE1NDVI, RE2NDVI

for prediction: carbon sequestered per meter of each sample canopy zone are the dependent varialbe (predicted outcomes). manually verifed with tape measures. 

Correlations are weak. R^2 in the range of (0, 0.22), p-value

Q: Correlation between different satellite images? 

SkySat images, pixel-based classification versus object-based delineation 

orthorectified and pansharpended, 

surface reflectance

Q: How training data was obtained or generated? 

Q: satellite images have trouble with height of trees which is important factor on biomass. 

UTC trees are smaller and younger than many other areas at Chattanooga. 

In remote sensing and GIS, a raster is a type of digital image that is made up of a grid of rectangular pixels. Each pixel in the grid contains a value that represents some aspect of a geographic feature, such as its elevation, temperature or land-cover type. Rasters are used for various kinds of analyses and modeling in environmental studies, geography, ecology and other fields.

Q: How is accuracy assessment achieved? 

Producer's accuracy, also known as user's accuracy, is a metric used in classification tasks to evaluate the accuracy of the positive predictions made by a model. It is defined as the proportion of true positive predictions (correctly predicted positive samples) out of all positive predictions made by the model. In other words, it is the probability that a positive prediction made by the model is actually correct.

Producer's accuracy = True Positives / (True Positives + False Positives)

This metric is useful in situations where the focus is on predicting positive samples accurately, such as in medical diagnosis where false positives can lead to unnecessary treatments or tests

The text discusses using allometric equations and biomass expansion factors to model sequestered carbon in urban tree canopies without the need for destructive harvesting. Remote sensing, specifically optical sensors, can be used to analyze vegetation and generate vegetation indices like NDVI and GNDVI, which can be used to estimate sequestered carbon. The study focuses on using Landsat imagery to analyze the changes in forest cover and urbanization in Chattanooga, TN from 1984 to 2021 with a 5-year interval. The researchers obtained 10 scenes during June and July for this study.

Only June and July data were used. 

Data usage: large image storage. 

LandSat is the oldest program, with coverage in Asia. can be used for eco habit study.