Statistical Stability and Spatial Instability in Mapping Forest Tree Species by Comparing 9 Years of Satellite Image Time Series

Abstract

Mapping forest composition using multiseasonal optical time series remains a challenge. Highly contrasted results are reported from one study to another suggesting that drivers of classification errors are still under-explored. We evaluated the performances of single-year Formosat-2 time series to discriminate tree species in temperate forests in France and investigated how predictions vary statistically and spatially across multiple years. Our objective was to better estimate the impact of spatial autocorrelation in the validation data on measurement accuracy and to understand which drivers in the time series are responsible for classification errors. The experiments were based on 10 Formosat-2 image time series irregularly acquired during the seasonal vegetation cycle from 2006 to 2014. Due to lot of clouds in the year 2006, an alternative 2006 time series using only cloud-free images has been added. Thirteen tree species were classified in each single-year dataset based on the Support Vector Machine (SVM) algorithm. The performances were assessed using a spatial leave-one-out cross validation (SLOO-CV) strategy, thereby guaranteeing full independence of the validation samples, and compared with standard non-spatial leave-one-out cross-validation (LOO-CV). The results show relatively close statistical performances from one year to the next despite the differences between the annual time series. Good agreements between years were observed in monospecific tree plantations of broadleaf species versus high disparity in other forests composed of different species. A strong positive bias in the accuracy assessment (up to 0.4 of Overall Accuracy (OA)) was also found when spatial dependence in the validation data was not removed. Using the SLOO-CV approach, the average OA values per year ranged from 0.48 for 2006 to 0.60 for 2013, which satisfactorily represents the spatial instability of species prediction between years.

Spatial–Temporal Vegetation Dynamics and Their Relationships with Climatic, Anthropogenic, and Hydrological Factors in the Amur River Basin

“…x are the values of variable x (vegetation indexes) at location i and location j, respectively. x is the average value of x, n is the total number of locations, and ij w represents the spatial weight [4,44,81,82].…”

“…As a major component of terrestrial ecosystems, vegetation plays an important role in material cycling and energy flows, and provides irreplaceable service functions that maintain the wellbeing of our planet and all the creatures that inhabit it. These function services include food provision, climate regulation, carbon sequestration, timber production, biodiversity preservation, and soil protection [1][2][3][4][5][6][7][8]. Vegetation growth affects the ecological balance, the terrestrial carbon cycle, water circulation, and other biochemical processes [2,9].…”

Mapping Floristic Patterns of Trees in Peruvian Amazonia Using Remote Sensing and Machine Learning

“…In each case, one fold was used as the validation set and the other folds as the calibration set. In the spatially constrained version, the inventory plots were divided into five latitudinally non-overlapping folds in order to reduce the possible effects of spatial autocorrelation [66][67][68][69][70]. This can be expected to give more conservative estimates of model performance.…”

Museo ToolBox: A Python library for remote sensing including a new way to handle rasters.

“…As stated by Olofsson et al (2014) "stratified random sampling is a practical design that satisfies the basic accuracy assessment objectives and most of the desirable design criteria". For spatial cross-validation, see Karasiak et al (2019) inspired by Roberts et al (2017).…”