CASA model validation
For this latest CASA model application, a comparison of observed NPP (n = 1927) from field based measurements to predicted annual values from the CASA model was made to provide validation of terrestrial NPP predictions across all ecosystem types. Observed NPP values were compiled for the Ecosystem Model-Data Intercomparison (EMDI) activity by the Global Primary Productivity Data Initiative (GPPDI) working groups of the International Geosphere Biosphere Program Data and Information System (IGBP-DIS; Olson et al., 1997). Monthly MODIS EVI inputs resulted in a highly significant correlation (R
2 = 0.90) and a close 1:1 match of observed to CASA predicted NPP values, with the year 2001 selected as an example (Fig. 1).
In this comparison to observed NPP, the CASA model was also tested for sensitivity to the vegetation index monthly time series as well by driving the NPP algorithm separately with either MODIS-EVI or MODIS-FPAR monthly inputs, holding climate inputs constant. A lower level of saturation in the low-to-medium range of plant production estimated from CASA modeling with EVI inputs compared to FPAR inputs was discovered by comparison of the two scatter plots over the range of annual NPP from 100 to 300 g C m−2 yr−1 (Fig. 1). Not only did EVI result in less overall scatter in the predicted versus observed plot (i.e., R
2 = 0.81 using MODIS FPAR inputs), the match to observed NPP in the high global range (of greater than 1000 g C m−2 yr−1 ) was markedly more consistent with EVI compared to MODIS FPAR inputs.
CASA NPP using EVI inputs was next compared across the tropical forest zones to a literature review of NPP estimates by Clark et al. (2001) from old-growth tropical forest study sites that attempted to create a consistent data set on NPP for these primary (undisturbed) ecosystems. Because NPP is composed of both aboveground and belowground forest production, upper and lower bounds around total NPP were reported by Clark et al. (2001) expressly to serve as benchmarks for validating biogeochemical models for this biome. The results of this compilation showed 58 % (22 out of 38) of sites estimated with mean annual NPP lower than 1000 g C m−2 yr−1, and another 26 % (10 sites) were estimated with mean annual NPP lower than 1200 g C m−2 yr−1 (Fig. 2). The comparison with the distribution of CASA NPP across the tropical forest zones indicated that large tracts (> 6 km2) of undisturbed rainforest with annual NPP higher than 1000 g C m−2 yr−1 (10 tons C ha−1) were nearly undetectable at the global scale. Despite the fact that hundreds of CASA annual NPP values of greater than 1000 g C m−2 yr−1 were predicted at isolated locations across all tropical forest regions of the world (indicating that the model has no built-in upper limit for NPP in its formulation), these high rainforest NPP cases (as reported by Clark et al., 2001) were extreme outliers in the global model distribution and hence do not appear as significant totals in Fig. 2.
By way of additional model validation in the tropical zone, comparison of CASA seasonal NPP against measured totals from the Large-scale Biosphere-Atmosphere Experiment in Amazonia (LBA) showed close agreement at the Tapajos (Pará) forest experimental site (Potter et al., 2009). At the ZF2 Manaus forest site, Chambers et al. (2004) directly measured respiration rates from live leaf, live wood, and forest soil surfaces to derive an indirect NPP flux estimate of 900 g C m−2 yr−1. Annual NPP from CASA for this general area (2.5 ° S lat, 60 ° W lon) around Manaus in Brazil varied between 782 and 871 g C m−2 yr−1 between 2000 and 2004.
Seasonal validation of CASA monthly NPP predictions from the MOD13C2 EVI data values closest to the AmeriFlux tower location was conducted by comparison to eddy-correlation monthly estimates of the corresponding NPP fluxes. It should be noted that the monthly MODIS EVI values in practically every grid cell of the global CASA model will be influenced by periodic land cover disturbances and areas of sparse vegetation cover, including development, roads, water bodies, and other natural features. It was expected, therefore, that CASA model NPP flux predictions would be systematically lower than tower measurements of these carbon fluxes, since tower footprints tend to be far less affected by wildfire and other disturbances, compared for instance to the surrounding MODIS grid cell area in which they are located.
A total of eight Ameriflux tower sites were found to meet the criteria cited in the methods section above for comparison to CASA model NPP predictions (Fig. 3). CASA model predictions closely followed the seasonal timing of Ameriflux tower measurements at each site, with a linear regression correlation coefficient of R
2 = 0.77 for all sites combined (Fig. 4). As expected, CASA predicted NPP underestimated by a modest level of 7 % compared to all the monthly NPP measured from the Ameriflux towers (at the setting of 40 % annual GPP carbon flux).
Predicted global NPP patterns
Predicted terrestrial NPP for the globe in 2009 was 50.05 Pg C, a total carbon flux in the middle of the range of previous vegetation NPP predictions of between 44 and 66 Pg C per year for the period 1982–1998 (Cramer et al., 1999). We estimate that global terrestrial NPP increased by +0.14 Pg C over the time period of 2000 to 2009, due almost entirely to a strong upward trend in the Northern Hemisphere (Fig. 5). Annual NPP was predicted to have increased between the years 2000 and 2007 in the regions of high-latitude (> 50° N) North America and Eurasia, and also in South Asia, West and Central Africa, and the western Amazon (Figs. 5 and 6).
This upward trend in high-latitude NPP was controlled by a combination of rapidly warming temperatures from 2004 to 2005 (Zhao and Running, 2010), and by elevated MODIS EVI patterns over the same period. Periodic declines in regional NPP were predicted for the southern Untied States, the southern Amazon, western Europe, southern and eastern Africa, and Australia (Fig. 7); the timing of negative NPP anomalies in each of these regions was associated with severe droughts and, in some cases, extreme heat waves (World Meteorological Organization, 2001 2009).
When global NPP predictions were broken down into 30° latitude zones, monthly air temperature was found to be highly correlated (R
2 > 0.9) with seasonal increases and decreases in NPP at latitudes between 30° N and 90° N (Table 1). Monthly precipitation was found to be more closely correlated (R
2 > 0.8) with seasonal increases and decreases in NPP than was air temperature at latitudes between 30° N and 30° S, whereas EVI was closely correlated (R
2 > 0.7) with monthly NPP in all latitude zones. Correlations between monthly EVI anomalies (2000–2009) and predicted monthly NPP anomalies were significant in all latitude zones as well, which implies that, across latitiude zones, interannual NPP variations were most strongly controlled by EVI inputs, compared to short-term variations in air temperature or precipitation.
Nonetheless, there were many instances of severe drought affecting terrestrial NPP on local-to-regional scales across the globe during the period of 2000 to 2008, mainly in areas of the central North America, Africa, Brazil, and China (Fig. 7). Beginning with major droughts in Brazil, the Horn of Africa, the Middle East, Central and South Asia, and China in 2000 and 2001, these events were followed by most of North America, southern Africa, and Australia experiencing record low precipitation amounts in 2002, 2003, and 2004. Large areas of Europe, southern Africa, Brazil, and Paraguay were affected by severe droughts in 2005. From 2006 though 2008, much of the United States, eastern and southern Africa, China, and Australia experienced continued deficits of precipitation.
Drought in the Amazon basin approached unprecedented levels in 2009 (and into 2010; Lewis et al., 2011), and the impact on NPP was evident in the CASA model prediction of 2009 anomalies in excess of −100 g C m−2 yr−1 in the central and western portions of the basin (Fig. 7). This was seen in contrast to the impacts of the severe drought of 2005, which was predicted to have impacted tropical forest NPP (with anomalies on the order of −20 g C m−2 yr−1) mainly in the eastern and in more isolated southwestern portions of the Brazilian Amazon.
A notable global difference of our CASA model estimates of NPP from those of previous terrestrial carbon studies (Cramer et al., 1999) was the magnitude of tropical forest NPP. Whereas these previously cited estimates of annual NPP across rainforest study sites that were largely protected from human disturbance have commonly exceeded 900 g C m−2 yr−1 (Clark et al., 2001), our CASA model for 2009 estimated an average NPP of 840 g C m−2 yr−1 in the South American tropical region, an average of 805 g C m−2 yr−1 in the African tropical region, and an average of 847 g C m−2 yr−1 in the South East Asian tropical region. Although hundreds of annual NPP values of greater than 1000 g C m−2 yr−1 were estimated as well at isolated locations across all three tropical forest regions of the world, the impact of deforestation and replacement of forests by lower production tropical agricultural systems has been widely captured by the MODIS EVI inputs to CASA since the year 2000.
Across the tropical rainforest zones, CASA NPP in 2009 totaled to 13.7 Pg C yr−1, accounting for nearly one-third (27 %) of global terrestrial NPP. Of the three major tropical rainforest regions, South America (80–43°W) accounted for 47 % of the zonal NPP total, whereas Africa (13.5–40°E) and Asia (73.5–162.5°E) regions together accounted for 45 % of the zonal NPP total (Table 2).