The use of precipitation intensity in estimating gross primary production in four northern grasslands

Abstract

Remote sensing is a useful tool for the estimation of gross primary production (GPP) in terrestrial ecosystems at regional to global scales. One limitation of remote sensing based GPP models is the inappropriate characterizing of precipitation impacts. In this study, we showed positive relationship between the monthly flux-measured GPP of four grasslands ecosystems and the precipitation intensity, which was calculated from dividing the monthly sums of precipitation by the half-hourly precipitation frequency. Suggested by this finding, two remote sensing based GPP models, i.e. the greenness and radiation model (GR) and the temperature and greenness (TG) model, were selected to test the potential of incorporating this precipitation intensity for the estimation of monthly GPP. A scaled precipitation intensity was proposed by normalizing a multi-year maximum precipitation intensity, considering its dynamical ranges across sites and regions. Results indicated that by adding of this scalar, the revised models can provide better monthly GPP estimates with average 10% improvements in precisions compared to their original outputs. A further analysis showed that such better performances of the revised models can be attributed to the positive relationship between precipitation intensity and the absorbed photosynthetically active radiation (APAR). However, no evident response has been observed on the light use efficiency (LUE), indicating the LUE and precipitation intensity relationship may differ across species and ecoregions. To the best of our knowledge, this is the first report of the potential use of precipitation intensity in the remote sensing based GPP models and it will be useful for the development of future models that can better predict GPP in the context of future precipitation regimes.

Introduction

Terrestrial ecosystems play a dynamic role in the global carbon (C) cycle as the carbon balance of terrestrial ecosystems is highly sensitive to climate changes, such as the inter-annual variations of precipitation regimes and increased surface temperatures as a result of the increase in atmospheric greenhouse gases (Beer et al., 2010). The increase in temperature and the elevated CO₂ in the atmosphere have been demonstrated to have great effects on terrestrial ecosystem production (Norby et al., 2005, Zhao and Running, 2010). Precipitation, on the contrary, has been suggested to have a more profound impact on ecosystem dynamics, especially in arid and semiarid environments (Weltzin et al., 2003). Changes in global and regional precipitation regimes are expected to have ramifications for the distribution, structure, and diversity of plants (Easterling et al., 2000). Although there are uncertainties for the feedbacks between annual precipitation variability and the aboveground net primary production (ANPP), it is suggested that precipitation patterns can alter the vegetation production and the responses may differ across biomes (Fang et al., 2001, Knapp and Smith, 2001).

Remote sensing has been an important tool for the estimation of gross primary production (GPP) at large spatial scales. Most of these models are based on the capturing of spectral characteristics of vegetation that are correlated to the biomass production. Specifically, the vegetation index (VI) are widely used in such models, providing an indicator of either the light use efficiency (LUE) (Garbulsky et al., 2011) or the faction of the absorbed photosynthetically active radiation (f_APAR) (Xiao et al., 2004). For example, the temperature and greenness (TG) model, derived by Sims et al. (2008), utilizes a combination of the enhanced vegetation index (EVI, Huete et al., 2002) and the land surface temperature (LST) in estimating GPP across different biomes. A total chlorophyll based model using the product of the normalized difference vegetation index (NDVI, Rouse et al., 1974) and incoming photosynthetically active radiation (PAR) also shows promising results for estimating GPP in crops (Gitelson et al., 2006, Wu et al., 2009) and forest landscapes (Wu et al., 2010). A common limitation of these remote sensing based GPP models is the inappropriate characterizing of the meteorological factors (e.g., temperature, precipitation) that can greatly affect LUE, leading to the largest uncertainty constrains the application of these models globally (Mu et al., 2011, Zhao et al., 2006). The reason is that the real-time LUE would change dramatically across seasons and between vegetation types that would be a function of factors as chlorophyll content, light, water, temperature (Hilker et al., 2008, Sims et al., 2008, Zhao et al., 2006) and thus a single vegetation index may have limited ability in LUE estimation, especially in extreme drought conditions (Samanta et al., 2010).

While impact on changes in the surface temperature has been considered in many GPP models (e.g., Coops et al., 2005, Sims et al., 2008, Xiao et al., 2004), little attention and result have been reported on views of precipitation regimes in such remote sensing based GPP models. Furthermore, precipitation regimes are predicted to become more variable with more extreme rainfall events punctuated by longer intervening dry periods. Manipulated experiments have demonstrated that these changes in patterns of precipitation can alter vegetation production (Heisler-White et al., 2008, Thomey et al., 2011). Therefore, it is urgent and necessary to improve those GPP models with incorporation of precipitation regimes, especially in consideration of the future regional and global climate changes (Paiva et al., 2011). Here we reported an analysis of the potential use of the precipitation intensity in remote sensing based GPP models. Satellite observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) images and multi-year flux measurements were used to estimate monthly GPP in four northern grasslands. The objectives are (1) to explore the potential for GPP estimation by incorporation of precipitation intensity, (2) to give analyses of reasons for the better performance of the revised models. These results will be useful for the development of future GPP models based on remote sensing observations and climate variables.