NationalMappingForestAttributes

R and Python scripts for the Postdoc project (collaboration between UBC's Integrated Remote Sensing Studio and the Canadian Forest Service).

Goal

Develop machine learning models (Random Forests, imputation methods) to relate time-series of Landsat images and LiDAR data to ultimately map forest attributes across Canada over a period of 30+ years, improving forest monitoring and management practices.

Approach and main results

Single year map

Multitemporal extension

Script description

• build_BorealMaps_YAI_2010.py: Build MXD with the 10 forest attributes maps for 2010 masked on the boreal

• build_Pyramids_UTM_maps.py: Build pyramids for all forest attributes maps in each UTM zone saved in 'E:\NTEMS\UTM_results'

• Build_save_parameters.R: Set global parameters and base directories for the suite of R scripts starting with 'prog0_lidar_plots_sampling_auto_GM.R'

• exportMaps_YAI_2010.py: Export to PNG the 2010 maps of some specified attributes of interest zoomed to a given bookmark

• exportTimeSeries_CAN_YAI_1984_2012.py: Export to PNG the time-series of specified attributes of interest zoomed on bookmarks across Canada (maps of 2nd paper)

• exportTimeSeries_elevp95_RF_1984_2012_YAI_2010.py: Export to PNG the time-series of some specified attributes of interest zoomed to a given bookmark

• Functions_NatMapping_Python.py: Define Python functions to make available to all py scripts

• Functions_NatMapping_R.R: Define R functions to make available to all R scripts

• merge_transects.py: Merge into a single shp all the individual shps containing the transects in each UTM zone

• prog0_lidar_plots_sampling.R: Define sampling grid over LiDAR transect

• prog0a_check_LiDARdata_row_duplicates.R: Check for row duplicates in all LiDAR csv files for all UTM zones

• prog0b_create_shp_from_csv.R: Create shp from CSV files and save new CSV files (lidar and forest attributes) with new unique_id for new LiDAR data from BC

• prog1_training_validation_selection.R: LiDAR training and validation plot selection

• prog1a_volume_biomass_check_recompute.R: check the issue with incoherence of Vol and Biomass values between Boreal/Non-boreal ecozones. Wrong equations have been used and this script recomputes the correct values on the transect for these 2 attributes.

• prog2_lidar_plots_footprint_extract_exvars.R: Extract values for all explanatory rasters spatially coincident with sampled lidar plot polygons for both training and validation plots

• prog3_model_selection.R: Descriptive statistics and plots, variable selection and Random Forest prediction/imputation

• prog4a_map_donor_plot_distance.py: Computes the distance from each pixel to its donor plot (in Lat/Long degree units) and also saves Lat and Long of donor plot

• prog4_check_maps.R: Check that predicted values on maps match those predicted on plots

• prog5a_rename_maps.py: renames maps after correction of biomass and volume values

• prog5_ecozone_stats.py: Compute descriptive stats by ecozone over the entire boreal for all the mapped attributes

• prog5_ecozone_stats_plots.R: Produce boxplots based on the descriptive stats extracted with 'prog5_ecozone_stats.py'

• prog6a_predict_on_random_pix.R: Predict forest attributes through time for randomly sampled pixels whose predictors are provided in a CSV file.

• prog6b_predict_on_val_pix.R: Predict forest attributes through time for validation pixels whose predictors are provided in a CSV file.

• prog6_temporal_analysis.R: Plot time-series of mapped attributes with boxplots of an ensemble of pixels sampled over large areas or with lines for individual pixel