CAFRI Labs: Big Tune Map Accuracy v2 - Training

Lucas Johnson

Description

Riemann et. al.

Previous version

Branch used to produce this document: big-tune-training

Coverages included

NYSGPO_WarrenWashingtonEssex_2015
NYSGPO_CayugaOswego_2018
NYSGPO_AlleganySteuben_2016
USGS_3County2014

FIA plot inclusion criteria

Training plots only
Tier 1 and 2 plots only.
All subplots must be sampled.

Masking

LCMAP Collection 1.1. data was used to develop masks for each LiDAR coverage included in the assessment, matching LCMAP products to LiDAR-AGB surfaces by year. E.g. LCMAP’s 2014 surface was used to mask the LiDAR-AGB surface for USGS_3County2014.

NOTE: Masked AGB pixels are set to 0, and included in the agreement assessment just as any other AGB pixel.

Masked classes:

1: Developed
5: Water

Agreement Statistics (SI corrected)

Table 1: Plot:Pixel Comparison; n = 226; Avg Plots Per Hex: NA
MODEL	MBE	RMSE	MAE	R2	KS	AC	ACs	ACu
RF	-1.31	38.25	27.48	0.73	0.11	0.66	0.97	0.69
GBM	-1.60	38.76	27.86	0.73	0.11	0.65	0.96	0.68
SVM	-1.37	33.55	22.57	0.80	0.10	0.75	0.98	0.77
LINMOD	-1.09	36.75	26.03	0.76	0.12	0.69	0.97	0.72
RMSE	-1.43	36.16	25.38	0.76	0.12	0.70	0.97	0.73

Table 2: 8660 Ha Hex; n = 170; Avg Plots Per Hex: 1.33
MODEL	MBE	RMSE	MAE	R2	KS	AC	ACs	ACu
RF	-0.22	32.66	24.44	0.77	0.11	0.71	0.98	0.74
GBM	-0.42	32.98	24.60	0.77	0.10	0.71	0.97	0.73
SVM	0.04	27.61	19.55	0.84	0.09	0.81	0.99	0.82
LINMOD	0.02	31.11	22.96	0.80	0.11	0.74	0.98	0.76
RMSE	-0.20	30.42	22.33	0.81	0.10	0.76	0.98	0.78

Table 3: 78100 Ha Hex; n = 48; Avg Plots Per Hex: 4.71
MODEL	MBE	RMSE	MAE	R2	KS	AC	ACs	ACu
RF	1.15	24.07	18.78	0.79	0.08	0.73	0.97	0.77
GBM	1.05	24.30	19.06	0.79	0.10	0.73	0.97	0.76
SVM	-0.04	22.17	16.88	0.83	0.10	0.76	0.96	0.80
LINMOD	0.97	23.48	18.24	0.80	0.08	0.75	0.97	0.78
RMSE	0.73	23.19	18.09	0.81	0.10	0.75	0.96	0.78

Table 4: 216500 Ha Hex; n = 26; Avg Plots Per Hex: 8.69
MODEL	MBE	RMSE	MAE	R2	KS	AC	ACs	ACu
RF	2.79	15.43	13.05	0.90	0.15	0.88	0.98	0.91
GBM	2.61	16.01	13.91	0.89	0.15	0.87	0.98	0.89
SVM	2.64	14.19	11.49	0.91	0.15	0.91	0.99	0.92
LINMOD	2.67	14.76	12.53	0.91	0.15	0.90	0.98	0.91
RMSE	2.68	14.92	12.81	0.91	0.15	0.89	0.98	0.91

ECDF Comparisons Across Scales - FIA Plots vs Mapped (SI corrected)

1:1 and GMFR Lines Across Scales - FIA Plots vs Mapped (SI corrected)

Aggregate AGB Estimates Across Scales (Uncorrected)

Spatial/Distribution Patterns of Local Differences (SI Corrected)

Patterns of Local Variability (SI Corrected)

Corrections

If you see mistakes or want to suggest changes, please create an issue on the source repository.

Citation

For attribution, please cite this work as

Johnson (2021, May 11). CAFRI Labs: Big Tune Map Accuracy v2 - Training. Retrieved from https://cafri-labs.github.io/acceptable-growing-stock/posts/big-tune-training-map-accuracy/

BibTeX citation

@misc{johnson2021big,
  author = {Johnson, Lucas},
  title = {CAFRI Labs: Big Tune Map Accuracy v2 - Training},
  url = {https://cafri-labs.github.io/acceptable-growing-stock/posts/big-tune-training-map-accuracy/},
  year = {2021}
}