VULNERABILITY OF VEGETATION IN PARTS OF HIMALAYAS AND DYNAMIC GLOBAL VEGETATION MODELLING ( DGVM ) – STUDY USING VNIR AND THERMAL RESPONSES OF MODIS TIME SERIES DATA

Vegetation responses to changing climate patterns need to be understood to devise adaptation strategy for a sustainable development, especially in the light of increasing climate related vulnerability. Dynamic Global Vegetation Models(DGVM) have the capacity and scope to develop understanding in this regard, due to their ability in simulating plant-vegetation-climate processes incorporating bioclimatic variables. However, prior to take up modelling using a spatially explicit DGVM, it may be imminent to prioritize the area for vulnerable contexts, so as to calibrate and validate the model optimally. Spatially explicit DGVMs require site level observations at canopy and leaf level/soil strata level for parametrization and implementation. Satellite data in VNIR and thermal regimes provide scope to understand the responses of various vegetation categories and enable to set up baseline addressing the foci of change as regions of vulnerability. Study carried out Western Himalayan transect using MODIS enhanced vegetation index and land surface temperature illustrates potential to differentiate areas that can be vulnerable due to warming trends disturbing cold to warm season energy level transition. Relations of these indices were studied in different vegetation categories and modelled spatially to derive potential vulnerable zones. Many sites showed high vulnerability while some sites showed distinct resilient behaviour by showing increase in EVI during warming periods. Potential zones were studied further using a spatially explicit Dynamic Global Vegetation Model for site level understanding. DGVM results in terms of biomass and carbon were studied to understand the trends in the vulnerable and resilient sites. Detailed characterisation of DGVM based modelling is underway to further diagnose the vulnerability contexts.


INTRODUCTION
Global warming is influencing vegetation growth and phenology (Zhang et al, 2004Parmesan & Yohe, 2003, Myneni et al. 1997, Cleland 2007).Vegetation responses to changing climate patterns need to be understood to devise adaptation strategy for a sustainable development, especially in the light of increasing climate related vulnerability.The responses can vary from cessation of normal phenological processes to increased physiological activity expressed, for instance, as improved vegetation vigour.Trends such as these may need to be confounded using other associated biophysical properties amenable at synoptic scale, which help to discern the levels of relative responses, in turn enabling decision making.Dynamic Global Vegetation Models(DGVM) have the capacity and scope to develop understanding of physiological responses , due to their ability in simulating plant-vegetation-climate processes incorporating bioclimatic variables.However, prior to take up modeling using a spatially explicit DGVM, it may be better to prioritize the area for vulnerable contexts, so as to calibrate and validate the model optimally.Understanding the disturbance in ecological contexts at global scale using combination of MODIS based EVI and LST parameters have provided insights in to collapse of leaf area due to sudden or gradual events (Mildrexler et al 2007, Mildrexler, 2009) and provide scope for applying in determining vulnerability.Regional scale DGVMs might present limitation in terms of attempting site scale calibration, which can be especially true for complex tropical forests, comprising of several plant functional types, yet to be included in to modelling.Downscaling regional DGVM operation to site level involving spatially explicit models may require alternate approach to know the vulnerability prevalent across a region.Such knowledge can help to validate these models more robustly.Satellite data in VNIR and thermal regimes provide scope to understand the responses of various vegetation categories and enable to set up baseline addressing the foci or hotspots of change as regions of vulnerability.
Theory involving thermal energy, its effect on ecological development as well as role of energy in functioning of selforganising systems and their organising ability are perceived to follow second law of thermodynamics (Kay 2000).Opportunity available with MODIS based Land Surface Temperature parameter provides scope for studying response of vegetation systems in terms of energy behaviour in comparison to vegetation condition as expressed by EVI.Land Surface Temperature has been derived using various fractions of electromagnetic spectrum (emissivity in bands 31 and 32 are estimated from land cover types, atmospheric column water vapor and lower boundary air surface temperature) as well as ground based relations.Surface temperature (which is the common product available from MODIS free online source from TIR remote sensing), Ts is addition of Air temperature and a ratio (ratio of aerodynamic resistance at boundary layer to specific heat of air weighted by a factor).Factor is difference between net all wave radiation at the surface and Latent heat flux i.e evapo-transpiration .Though it is a relatively complex set of equations to arrive at T s it is worth understanding keeping in view of its ability to understand how vegetation responses are deviating from the normal in tune with warming trends.The equation form of the concept above is Where R n -Net all wave radiation at the surface, R b -Boundary layer aerodynamic resistance, Cߩ -Emissivity and LE -Latent Heat Flux (Evapotranspiration) .Latent Heat Flux is affected by plant, air and radiation properties like a) resistance at canopy by stomata and LAI, b) resistance offered by dynamics of air movement, c) vapour density deficit of the air (packing of water molecules), d) specific heat of the air (how much energy air can retain), e) density of the air, f) slope of saturation vapourpressure relationship, g) flux of energy in to or out of the surface (soil, leaf, water etc) and h) net all wave radiation at surface.Enhanced Vegetation Index employed in the study is an optimized version standardized in MODIS sensor, and is sensitive to high biomass regions.EVI helps to monitor vegetation better through de-coupling of canopy background signal and reduced atmospheric influences .
Potential vulnerability as a satellite derived understanding attempted here, derives rationale from the fact that any increase in land surface temperature with concomitant increase in foliage manifestation retains scope for eco-physiological resilience, while otherwise, situation may push vegetation in to sub-productive situation due to linked eco-physiological limitations causing disturbed phenology.

Study Area
A study window covering sufficient biogeographic variation corresponding to diverse phenological categories of vegetation in Western Himalayas was chosen between geographic extent of 30 20 55. 68

Satellite Data Used
Datasets from MODIS EVI (Enhanced Vegetation Index) and LST (Land Surface Temperature) composites (16 and 8 day respectively) for area covering vegetation categories from subtropical to cold arid alpine in Himachal Prdesh, India were analysed.EVI and LST have ability to represent plant performance/growth and potential stress respectively.While EVI compensates better for saturation issues of LAI, LST integrates extrinsic/intrinsic factors of land cover warming.Selection of the data was done to represent distinct seasons each in a normal and anomalous year, after studying temporal responses in 2000-2009 time series data.Data set for 2009 was known thermally anomalous year and 2006 a normal year, as defined by IMD.

Homogenous Vegetation Strata:
Homogeneity of vegetation is a percept linked to scale of observation.However, for a current inquiry, satellite based pixels accounting for the continuity of land cover within 3X3 pixels dimension is construed so, to facilitate coarse resolution (1X1km) correspondence of 24 m vegetation type map.Hence vegetation type map was resampled to 120 m resolution (5 X 5 pixels) and visually best possible square window was marked as to correspond to HVS, assuming any discontinuity as part of the pixel only.Vegetation categories principally corresponded to forest and natural vegetation as defined under national database prepared for Biodiversity Characterisation Project (Sponsored by Department of Space and Biotechnology jointly).Under this national project more than one season satellite data was used to prepare forest and other natural cover maps in entire India.Database for the study area selected was prepared during Phase I of the study using IRS LISS III datasets of 1999-2000 period.Coarse grid database consisting of 120 m pixel was used to draw 'homogenous vegetation strata' transects across entire Country.The HVS transects had to adopt to the varying patch shape and sizes of 37 categories considered.Transects varied from 4 x 4 km to 1 X 1 km in size so as to sample at least one EVI/LST pixel (of 1000 m pixel size) and they corresponded to at least 1% of the cover.

Response of vegetation categories in EVI and LST
Euclidean space: Responses selected over each representative vegetation category were plotted as type-wise means across EVI-LST space for cold (January) and warm (April) months of normal (2006) and anomalous (2009) year.Major categories depicted differentiation and demonstrated the inherent potential of thermal and vegetation index combination to differentiate intended forest vegetation phenological categories.Comparison of normal and warm year distribution in terms of the displacement (highlighted by direction and magnitude of arrows empirically) within the Euclidean space (as depicted in Fig. 3) pointed out the potential of Euclidean distance that can be employed for discriminating responses.In order to track the significantly deviating pixels, values showing magnitude beyond the average magnitude (beyond two sigma) were selected, so that extrema of the movement is represented as anomaly.It was assumed that values performing within two-sigma Euclidean distance of EVI-LST space were not depicting any abnormal behavior hence were not candidates of being vulnerable/responsive to warming trends prevalent.Method was assessed in a spreadsheet for calibrating the representation of distinction and implemented in ERDAS Imagine 2010 suite using modeler functionality.
Euclidean distance between respective indices between each year was used for measuring the departures from normal quantum and direction of variation.Direct ratios between and within year Consequent to the delineation of vulnerable zones in the study region, attempt has been made to apply a dynamic vegetation growth model at specific instances of varied resilience.Realization by climatic researchers of the need to address the complexities of vegetation surface at finer scale, as well as influence of vegetation on climate trends, has greatly influenced the integration of different biophysical models towards coupling.Spatially Explict Individual Based Simulator (SEIB) -DGVM (Sato et al, 2007) (Ver.2.54 downloaded : Jul 2011, seibdgvm.com), complied using g95-MinGW was employed for developing understanding in sites showing vulnerability.Initial attempts were focused on ability of the model to synthesize extant vegetation at the site and develop further upon the same.Though we proposed the use of LPJ-GUESS Education module , its specific intent of being only education version made us to use this open source software.SEIB relies on a three dimensional interpretation of the real world vegetation with improved light interception and allocation routine.Its disturbance module also covers occurrence of fire at a critical biomass spill over.With 2.54 vegetation suited to field compatibility were found to be modeled.Model run was executed for 50 years at site showing vulnerability (Hanle) as well as at site showing resilient behavior (Sangla).SEIB model synthesized (Fig. 5) tall grass and boreal woodland PFTs in these two sites respectively which matches the expected vegetation on ground.The model results show a trend characteristic of grassland and woodland accumulating biomass over half century and without any loss of substantial carbon over any point of time, possibly due to cessation of growth.Emissions seem to go up in later part, possibly due to the quantum of carbon available at pool.Woodland seems to show dip in the growth in first half of 50 years and to catch up later for a continuous growth.However , since the climate pattern for 50 years is assumed to be as that of calibrated year sequence ( NCEP reanalysis available for modelling through SEIB link), results can be considered as inputs for calibrating model performance in test site.A collaboration is underway with institute involved in this region for calibrating the model performance.Detailed behaviour of the site would be studied using climate data adapted to SEIB from ongoing efforts using models like WRF.

CONCLUSION
Developing upon the approach of disturbance index using MODIS data, this study illustrates the potential of EVI in conjunction with LST to delineate pockets of vulnerability in the context of global scale vegetation responses.Such hot spots picked can serve as sites for applying dynamic vegetation models to understand vegetation behaviour and devise strategies of adaptation if required.Such trends especially in highly ecologically fragile set up of alpine and temperate bioclimatic setup are critical from the view point of loss of habitats and species vulnerability also.Spatially explicit vegetation models can precisely help to understand functional aspects of such pattern deviations and strengthen adaptation programmes.
N, 75 17 51.09E (Lower left), 33 21 04.23 N, 79 18 19.64E (Upper right) covering 482 columns and 362 lines of MODIS data.Such an extent enables wide range of responses possible across terrain and climate range.Window covered entire Himachal Pradesh as well as parts of Jammu and major parts of Southern Ladakh, corresponding to Hanle and Puga valleys (Fig 4).Study area covered following vegetation types, which are characteristic of most of the Himalyan vegetation transects.-Tropical ( Dry and Moist Deciduous ), Sub-tropical (Mixed Chir pines), Temperate (Conifer, Broadleaf), Deodar gregarious as well as tropical and Alpine grasslands .Region experiences a temperature range of -25 Deg C (Hanle) to 45 Deg C (Baddi, Himachal) and an average annual rainfall range of 200 -2000 mm.