Climate Change Projections of Terrestrial Primary Productivity over the Hindu Kush Himalayan Forests

Increasing atmospheric carbon dioxide concentration [CO2] caused by anthropogenic activities has triggered a requirement to predict the future impact of [CO2] on forests. The Hindu Kush Himalayan (HKH) region comprises a vast territory including forests, grasslands, farmlands and wetland ecosystems. In this study, the impacts of climate change and land use change on forest carbon fluxes and vegetation productivity are assessed for HKH using the Lund-Potsdam-Jena General Ecosystem Simulator (LPJ-GUESS). LPJ-GUESS simulations were driven by an 15 ensemble of three climate models participating in the CMIP5 (Coupled Model Intercomparison Project Phase 5) database. The modeled estimates of vegetation carbon (VegC) and terrestrial primary productivity were compared with observation-based estimates. Furthermore, we also explored the net biome productivity (NBP) and VegC over HKH for the period 1850-2100 under the future climate scenarios RCP2.6 and RCP8.5. A reduction is observed in modeled NBP and VegC from 1951-2005 primarily due to land use change. However, an increase in both NBP and 20 VegC is predicted under RCP2.6 and RCP8.5. The findings of the study have important implications for management of the HKH region and inform strategic decision making, land use planning and clarify policy concerns.

downloaded from "The Application for Extracting and Exploring Analysis Ready Samples (AρρEEARS)" website ("LP DAAC -AppEEARS".). Land cover (MOD12Q1) used in this study was downloaded from 145 files.ntsg.umt.edu/data/NTSG_Products/MOD17/GeoTIFF/MOD12Q1/ and was used for land cover stratification (Friedl et al., 2002). Land cover related to barren, water and urban were masked from LPJ-GUESS data in order to make it comparable with MOD17A3 data (i.e. identical spatial extent, land cover classes and number of grid cells).

Comparison between Observed and LPJ-GUESS estimations of VegC
Simulations forced by three CMIP5 Earth System Models (ESMs) of mean VegC from 1990-2015 were compared with the observed GEOCARBON dataset (Figure 2). The mean VegC of observed dataset was estimated to be 4.68 kg C m -2 . While the modeled VegC for HKH averages 1.92 kg C m -2 , 2.04 kg C m -2 and 2.11 kg C m -2 for simulations 155 forced by climate outputs from IPSL-CM5A-MR, MPI-ESM-LR and CCSM4 respectively. Most of the difference is found to be the southern regions of HKH A moderate agreement was found between the GEOCARBON and LPJ-GUESS VegC with a mean r 2 value of 0.44. Furthermore the simulations of the CMIP5 models and the observed estimations in the HKH region were compared according to land cover classes from MOD12Q1 ( Figure 3). There is a large overestimation of VegC in evergreen broadleaf forests. The mean GEOCARBON VegC was 7.73 kg C m -2 was on average, 2.68 kg C m -2 higher than LPJ-165 GUESS VegC for evergreen broadleaf forest. Overestimation was also observed in grasslands and open shrublands.
VegC for remaining forest types showed a lesser difference than 1.5 kg C m -2 . It is important to note that the simulation of VegC in not very sensitive to differences in the modelled climates from the CMIP5 models for the period from 1990-2015. https://doi.org/10.5194/esd-2020-84 Preprint. Discussion started: 26 November 2020 c Author(s) 2020. CC BY 4.0 License.

Evaluation of patterns of GPP and NPP from 2000-2010
The mean MODIS GPP for 2000-2010 was estimated to be 0.69 ± 0.26 kgC m -2 yr -2 . The GPP for IPSL-CM5A-MR, MPI-ESM-LR and CCSM4 was 0.84 ± 0.17 kgC m -2 yr -1 , 0.83 ± 0.16 kgC m -2 yr -1 and 0.88 ± 0.16 kgC m-2 yr -1 respectively ( Figure 4). The mean MODIS NPP was estimated to be 0.40 ± 0.16 kgC m-2 yr -1 and 0.43 ± 0.07 kgC m-2 yr -1 , 0.42 ± 0.07 kgC m-2 yr -1 , and 0.44 ± 0.07 kgC m-2 yr -1 for IPSL-CM5A-MR, MPI-ESM-LR and CCSM4 180 respectively ( Figure 4). Both of the spatial datasets are able to capture important features such as the low productive Himalayan tundra ecosystem in the north and high productive regions like the forests and croplands in lower parts of HKH region ( Figure 5 & 6). There was a moderate spatial agreement between the MODIS and modelled GPP with mean r 2 values of 0.54. However, there was a weak correlation between the satellite-derived and modelled NPP with

Projected Spatial Changes in the Pattern of NBP and VegC
The spatial maps for NBP shown in Figure 8 presents the averaged spatial NBP estimated for 1850-1950, 1951-2006 205 and 2006-2100 (RCP2.6 and RCP8.5) for CCSM4 respectively. The results of IPSL-CM5A-MR and MPI-ESM-LR are included in the supplementary information, as all three models showed a similar trend for both variables with minimal difference (Supplementary Figure S1 and Figure S2). The LPJ-GUESS mean NBP from 1850 to 1950 is 0.0011 kgC m -2 yr -1 and -0.0017 kgC m -2 yr -1 for 1951 to 2005. The simulations have shown a shift from carbon source to sink in both future scenarios with mean NBP of 0.0206 kg C m -2 yr -1 and 0.0466 kg C m -2 yr -1 for RCP2.6 210 and RCP8.5 respectively.  The estimated total NBP for HKH region for the three CMIP5 models are shown in Figure 10. The temporal trend of total net biome productivity for the three CMIP5 models. In 1951-2005 (with respect to past period) MPI-ESM-LR and CCSM4 show a decreasing trend of total NBP, however IPSL show an increase for that time period. The average 230 total NBP for RCP2.6 was estimated to be 55 kg C m -2 yr -1 and for RCP.8.5 it was 111 kg C m -2 yr -1 . All ESMs follow The total of VegC ( Figure 11) was estimated for the HKH region, for three different time period  2005 and for future scenario 2006-2100 (RCP2.6 and RCP8.5) respectively. Model estimates of total VegC in HKH terrestrial ecosystems have increased since the 2005 and will increase under both future climate scenarios. The total VegC (averaged for all models) was estimated to be 7400 kg C m -2 by 1950, and by 2100, it is projected to range to 6000 kg C m -2 under the RCP2.6 scenario and 7600 kg C m -2 under the RCP8.5. Spatial patterns show that the mean VegC ( Figure 12

Discussion
The global biomass datasets based on inventories and satellite observations have been recently available. For our first approach we compared the modelled simulations VegC and primary productivity with satellite observations. For 265 VegC, the observed dataset is a global biomass map from the GEOCARBON project, a product of aboveground biomass dataset for the year 2000. A moderate agreement was found between GEOCARBON and IPSL-CM5A-MR forced simulated VegC and low agreement was found when climate data was supplied by MPI-ESM-LR and CCSM4.
The difference between modelled and observed VegC may be attributed due to the differences in terms of the coverage of aboveground or belowground biomass of both datasets. The GEOCARBON dataset includes the spatial distribution It is important to note that the DGVMs including LPJ-GUESS and the MODIS algorithm do not share a common meteorological drivers and that might be the potential to bring out a weak to moderate correlation between the two datasets (Liu et al., 2018). Previous literature have also reported that DGVMs generally overestimate GPP. Yet most 285 of the researchers suggest that simulated GPP by DGVMs were neither overestimated nor underestimated, but the results are limited by number of observational or model abilities. For instance our modelled LPJ simulations have few important processes missing such as impact of nitrogen deposition (Li et al., 2016). The inconsistencies of primary productivity for EBF was also observed in various studies (Ardö, 2015;Garrigues et al., 2008). Areas affected by frequent cloud cover or atmospheric contamination may then show inconsistent estimates of vegetation productivity 290 using MOD17 (or any method based on satellite based observations).
The second approach was to explore the variability of NBP and VegC over HKH from 1850-2100 and how this greater increase in VegC and NBP was seen in RCP8.5, as rate of photosynthesis by terrestrial vegetation rises due to increase with atmospheric CO2 content leading to increased carbon uptake. Global scale study carried out by (Thompson et al., 2004) discussed that the CO2 fertilization could limit the global warming in the future scenario, however the nutrient limitations could weaken this effect. However, the version of LPJ-GUESS used in this study did not take account of nutrient limitations and assume nitrogen to be at an optimal level for the terrestrial vegetation. The 320 coupling of carbon and nitrogen cycles are becoming widely recognized as nitrogen dynamics have been incorporated into global C cycling model (Fleischer et al., 2015). However there is large variation in the net effect on NBP and VegC due to uncertainties arising from different climatic forcing from respective ESMs and its description of future climate (Ahlström et al., 2017).

Conclusion 325
The results of the study has indicated that HKH will act as a net sink of C. However, the extent to which it will remain a C sink is uncertain as the parameterizations in LPJ-GUESS are in the early stages of validation and implementation in climate models. Uncertainties and large variation was found between the observed and modelled datasets. It is important to note that as long as obtainability and access of meteorological data at a regional level and in situ validation data such as eddy covariance measurements and long-term ecological field assessments remain scarce, it can be 330 expected the representativity of vegetation carbon and vegetation productivity estimates for HKH to remain hard to evaluate and determine. The LPJ-GUESS simulations revealed that the NBP is projected to be higher in future scenario given that the LULCC remains stable. Furthermore VegC storage spatial and temporal analysis suggest that, for the RCP8.5 scenario, the CMIP5 ESMs produces, on average, a slightly higher VegC compared to the RCP2.6 attributing to the CO2 fertilization effect. It is predicted the region will act as net sink of C in the future scenario. Vegetation 335 fluxes can help to analyze the carbon storage patterns, however further studies are required to assess the effects of climatic changes and anthropogenic activities on the fragile ecosystems of the HKH for the establishment of policies to improve the livelihood of the local population and the overall carbon balance in the region.