Assessment of Regional Forest and Scrub Productivity Using a Coupled Vegetation Process Model with Remote Sensing

Abstract

The coupling of remote sensing predicted variables within vegetation process models is well suited to predicting carbon balances in vegetation communities at regional or national scales as spatial and temporal variations detectable from satellite sensors are likely to be related to broad‐scale environmental and disturbance patterns. Over the north and south islands of New Zealand we couple a simple vegetation process based model 3PG (Predicting Physiological Principals of Growth) with constraints provided by estimates of the fraction of photo‐synthetically absorbed radiation (fPAR) derived from the SPOT 4 satellite. Model predictions of national forest and scrub above ground Net Primary productivity (NPP_A) were compared to estimated values derived from previously published values at both individual plot locations and regions, stratified by vegetation type.

Comparison of the field estimated fraction of fPAR values across the range of vegetation types in the North and South Islands and SPOT 4 VEGETATION NDVI indices indicated the imagery explained 83% of the variation in fPAR. The 3‐PG predictions coupled with remote sensing observations of above ground stand NPP_A for 6 individual sites across both islands correlated well (adjusted r²=0.80, p < 0.001) with field estimates however the relationship had a significant offset with the model over predicting stem NPP_A of individual plots. A comparison of regional estimates of stem NPP_A also correlated well (adjusted r²=0.74, p < 0.001) again with model prediction, having a significant offset. A key discussion point is the accuracy of model verification, as scalable plot and regional estimates of forest and scrub NPP_A with which to verify the model are difficult to obtain and often unsuited for comparison. Despite this, the correspondence between values of predicted NPP_A by the coupled model and those estimated from conventional forestry measurement techniques indicates that the model provides a sound and useful framework for regional modeling of forest growth. The results also highlight that, from the model output, it is possible to identify the constraints imposed on growth by vapor pressure deficit, soil water balance and temperature.