Remote Sensing Estimation and Spatiotemporal Pattern Analysis of Terrestrial Net Ecosystem Productivity in China

Abstract

Net ecosystem productivity (NEP) plays an important role in understanding ecosystem function and the global carbon cycle. In this paper, the key parameters of the Carnegie Ames Stanford Approach (CASA) model, maximum light use efficiency (εmax), was optimized by using vegetation classification data. Then, the NEP was estimated by coupling the optimized CASA model, geostatistical model of soil respiration (GSMSR) and the soil respiration–soil heterotrophic respiration (Rs-Rh) relationship model. The ground observations from ChinaFLUX were used to verify the NEP estimation accuracy. The results showed that the R2 of the optimized CASA model increased from 0.411 to 0.774, and RMSE decreased from 21.425 gC·m−2·month−1 to 12.045 gC·m−2·month−1, indicating that optimizing CASA model by vegetation classification data was an effective method to improve the estimation accuracy of NEP. On this basis, the spatial and temporal distribution of NEP in China was analyzed. The research indicated that the monthly variation of NEP in China was a single peak curve with summer as the peak, which generally presented the pattern of southern region > northern region > Qinghai–Tibet region > northwest region. Furthermore, from 2001 to 2016, most regions of China showed a non-significant level upward trend, but main cropland (e.g., North China Plain and Northeast Plain) and some grassland (e.g., Ngari in Qinghai–Tibet Plateau and Xilin Gol League in Inner Mongolia) showed a non-significant-level downward trend. The study can deepen the understanding of the distribution of carbon sources/sinks in China, and provide a reference for regional carbon cycle research.

Estimation of European Terrestrial Ecosystem NEP Based on an Improved CASA Model

“…The NEP values of DBF, ENF, EBF, DNF, and MF are higher, while those of CRO, GRA, OSH, and WET are relatively lower. This is consistent with the study conducted by Liang et al [35] in China, which shows that different vegetation types have great differences in different geographical regions. The result further illustrates the necessity of classifying different vegetation types when using the CASA model to estimate vegetation productivity.…”

“…Then, we extracted the model estimated NEP value of 1 × 1 km pixels of each corresponding point of the observation station by GPS coordinates. Finally, the R 2 and RMSE are used as indicators to evaluate the accuracy between NEP estimates and measurements [35], [36] (see Fig. 1).…”

“…The results show that this method can overcome the problem of low estimation values and effectively improve the estimation accuracy of the model (R 2 increases from 0.252 to 0.403, and RMSE decreases from 84.557 gC•m −2 •month −1 to 64.466 gC•m −2 •month −1 ). In addition, the research of Liang et al [35] shows that the estimation accuracy of NPP of various ecosystems in China can be improved by using the ε max of various vegetation provided by Zhu et al [40] and the LUE values used are different from those in this article [9]. In addition, the research of Liang et al [35] shows that using ε max of various vegetations provided by Zhu et al [40] can improve the estimation accuracy of NPP in China, and the ε max values used are different from those in this article [9].…”

“…In addition, the research of Liang et al [35] shows that the estimation accuracy of NPP of various ecosystems in China can be improved by using the ε max of various vegetation provided by Zhu et al [40] and the LUE values used are different from those in this article [9]. In addition, the research of Liang et al [35] shows that using ε max of various vegetations provided by Zhu et al [40] can improve the estimation accuracy of NPP in China, and the ε max values used are different from those in this article [9]. This shows that determining the ε max value of CASA model needs to consider different vegetation types as well as different geographical regions.…”

“…However, this method requires daily average data products to be screened, which requires high temporal resolution temperature data products and is greatly affected by random weather changes. Therefore, it is generally contended in subsequent studies that this value should be taken as the temperature when the NDVI monthly change curve reaches the peak [10], [35], [47]. However, the peak value of NDVI monthly change corresponds to the period with the maximum accumulation of vegetation biomass, which is the balance point of vegetation growth and consumption, rather than the period with the fastest vegetation growth.…”