Click here to insert picture ECOSYSTEM HEALTH ASSESSMENT BASED ON REMOTE SENSING A CASE STUDY OF CA RIVER BASIN, VIETNAM Bao Quoc Tran MSc Thesis WM-WRM.16-22 Student number: 47074 April 2016 ECOSYSTEM HEALTH ASSESSMENT BASED ON REMOTE SENSING A CASE STUDY OF CA RIVER BASIN, VIETNAM Master of Science Thesis by Bao Quoc Tran Supervisors Prof Wim G.M Bastiaanssen Mentors Hans van der Kwast, PhD Tim Hessels, MSc Examination committee Prof Wim G.M Bastiaanssen Hans van der Kwast, PhD Tim Hessels, MSc Ir G.J Roerink (WUR-Alterra) This research is done for the partial fulfilment of requirements for the Master of Science degree at the UNESCO-IHE Institute for Water Education, Delft, the Netherlands Delft April 2016 Although the author and UNESCO-IHE Institute for Water Education have made every effort to ensure that the information in this thesis was correct at press time, the author and UNESCOIHE not assume and hereby diSLCaim any liability to any party for any loss, damage, or disruption caused by errors or omissions, whether such errors or omissions result from negligence, accident, or any other cause © Bao Quoc Tran 2016 This work is licensed under a Creative Commons Attribution-Non Commercial 4.0 International License Abstract The Ca river basin is the third biggest river basin in Vietnam, in which the upper part belongs to Laos With the World Biosphere Reserve Western Nghe An officially recognized by UNESCO in 2007, the up-to-date information about ecosystem and biodiversity, in particular, flora biodiversity is urgently needed for conservation strategies as well as policy making process in resource management This study aims to evaluate Skidmore et al (2015)’s proposed variables to assess the ecosystem health in the Ca river basin from remote sensing indices, namely leaf area index, soil moisture, net primary production, land use and fire occurrence with the hypothesis that this framework is generic for all ecosystems The remotely sensed imagery was retrieved from Landsat ETM+ SLC-off in March of three years, namely 2005, 2010 and 2015 In addition, a weighted scoring approach has been attempted to assess the vigor aspect of ecosystem health The results showed that the LAI was underestimated, which might imply that the function retrieving LAI from SAVI for all crops was not applicable in this study area In addition, the calculation of soil moisture should be taken into account the weather condition since they were estimated in a day only Accordingly, the fire occurrence map also pointed out some areas where the fire events happened at least twice in three time steps, which might be caused by slash and burn practices of local inhabitants to prepare for the next crop Related to ecosystem health assessment in 2010 and 2015 in comparison with 2005, which is considered a benchmark, the ecosystem of the study site in 2010 was moderate healthy and getting viable in 2015 On the other hand, Skidmore et al (2015)’s approach remains with technical and conceptual limitations Likewise, it is believed that an agreement on an optimal resolution, in terms of temporal, spatial and spectral resolutions, should be drawn among research communities to bridge the gaps between remote sensing experts and ecological users Furthermore, the weighted scoring approach should be integrated between professional biologists as well as the statistical data from the field to minimize the bias In the end, local calibration is a must since every ecosystem has its own characteristics and the algorithms might not be applied to all ecosystems Keywords: Ecosystem Health Assessment, remote sensing, weighted scoring approach i Acknowledgements I would like to thank Dr Hans van der Kwast for his support and many helpful contributions over the past five months His suggestions on how and where to find support for this study was a major contributing factor to its completion A special thanks to Tim Hessels who encouraged me and engaged his time in technical issues Together with intensive assignments and meetings almost every week, their requirements have been keeping me on track with the thesis working so that I could finish my thesis on time and engaged myself in scientific research related to remote sensing This paper would not have been possible without Professor Wim G.M Bastiaanssen, who first inspired me to this thesis topic from a lecture given in early last year and guided me with the theoretical concepts on remote sensing and biodiversity I would also like to acknowledge Adeline, my best friend and my classmate, for assisting with many questions that I had on English writing and discussion We did have a lot of memories when working in DOK, TU Delft library since morning until midnight with nice coffee and food A special thanks to Louis who is sharing with my all the sorrows we had when struggling with thesis writing We did have several stories to tell about life, about love, about tears, even about the relationship between the duck-canal network in Delft with biodiversity I will miss the time we travelled together to Berlin and worked hard with our thesis on the bus A special thanks to Jam, Mariel, Clara, Shabana, Saltana and other Water Management classmates who always take care of me and spent crazy time with me in last 18 months Last but not least, I would like to leave the last paragraph to give all my love to my family, who always encourages and are beside me unconditionally Finally, I would like to give a big hug to my boyfriend, Quốc Trạng, who did encourage me to get this scholarship in last two years and supports me with love and humor, smile and tears, strengths and efforts to overcome all the obstacles in my life I love you iii Table of Contents Abstract i Acknowledgements iii List of Figures ix List of Tables xi Abbreviations xiii Introduction 1.1 Background 1.2 Problem statement 1.3 Objectives 1.4 Hypothesis 1.5 Research questions 1.6 Study site background 1.7 Structure of the thesis Literature Review 2.1 Ecosystem Health Assessment (EHA) 9 2.2 Skidmore et al (2015)’s proposed biodiversity variables 10 2.3 Remote sensing indices as required inputs 12 2.3.1 Normalized Difference Vegetation Index (NDVI) 12 2.3.2 Soil Adjusted Vegetation Index (SAVI) 13 2.3.3 Normalized Burned Ratio (NBR) 14 2.3.4 Land surface temperature 15 2.4 Using indices to retrieve biodiversity variables 16 2.4.1 Leaf area index (LAI) 16 2.4.2 Leaf Nitrogen Content 16 2.4.3 Soil moisture 17 v 2.4.4 Land cover 18 2.4.5 Vegetation height 19 2.4.6 Burn severity levels 20 2.4.7 Vegetation phenology 21 2.4.8 Net primary production 22 2.4.9 Inundation 22 Case study site description 25 3.1 Location 25 3.2 Ecohydrological characteristics 26 3.2.1 Topography, geology and soils 26 3.2.2 Climate 26 3.2.3 Flora and fauna 27 3.2.4 Land use and land cover 29 3.2.5 Pressures, threats and current outlook 30 Methodology 33 4.1 Research strategy 33 4.2 Data collection, processing and analysis 35 4.2.1 Description of sensors available 35 4.2.2 Preprocessing Landsat ETM+ SLC-off 36 4.3 Retrieving remote sensing indices and biodiversity variables from Landsat ETM+ 39 4.3.1 Estimating land surface temperature 40 4.3.2 Soil moisture 42 4.4 Burn severity levels and fire occurrence 43 4.5 Zonal statistics 43 4.6 Ecosystem Health Assessment 44 Results 47 5.1 Derivation of biodiversity variables from remote sensing 47 5.1.1 Leaf Area Index (LAI) 47 5.1.2 Soil Moisture 51 vi 5.1.3 Burn severity levels and fire occurrence 55 5.1.4 Net primary production 57 5.2 Ecosystem Health Assessment using Weighted Scoring approach Discussion 60 65 6.1 Derivation of biodiversity variables from remotely sensed imagery 65 6.2 Limitations of Skidmore et al (2015)’s framework and suggested solutions for improvement 68 6.2.1 Technical limitations 68 6.2.2 Conceptual limitations 70 Conclusions and Recommendations 75 7.1 Conclusions 75 7.2 Recommendations 76 References 77 Appendices 95 Conversion DNs to TOA brightness temperature 95 Python Scripts 97 Net primary production and World Net Primary produtivity for major ecosystems 103 vii List of Figures Figure 1-1 Ecosystem services (Sheet 7) in Water Accounting Plus Framework Figure 1-2 The research flowchart Figure 2-1 NDVI and SAVI calculated from a Landsat TM5 image of south-western Idaho 13 Figure 2-2 Spectral response curves of vegetation and burned area 14 Figure 2-3 Flowchart of global vegetation classification logic 19 Figure 2-4 Conceptual representation of a forest standing indicating the relative positions of mean canopy height and scattering phase center height within a single SRTM resolution cell 20 Figure 3-1 Location and topographic map of the case study 25 Figure 3-2 Location of Phu Xai Lai Leng 26 Figure 3-3 Climatogram of study area, data at Vinh station (2013) 27 Figure 3-4 World Biosphere Reserve Western Nghe An 27 Figure 3-5 Representative flora and fauna in study area 28 Figure 3-6 The trend in land use structure from 2000 to 2013 29 Figure 3-7 Land Use map of Ca River Basin, Nghe An, Vietnam (2012) 30 Figure 4-1 The flowchart of retrieving biodiversity variables from remote sensing 34 Figure 4-2 The difference of with and without SLC in processing image 36 Figure 4-3 Inverse Distance Weight Interpolation based on weighted sample point distance (left) and Interpolated IDW surface from elevation vector points (right) 37 Figure 5-1 Leaf Area Index in three time steps (2005, 2010, 2015) 48 Figure 5-2 Mean LAI in three time steps per land use 48 Figure 5-3 Soil moisture in three time steps 51 Figure 5-4 Soil moisture per land use in three time steps 52 Figure 5-5 Burn severity levels of study area in three time steps 55 Figure 5-6 Frequency of fire occurrence from 2005 to 2015 56 Figure 5-7 Net primary production in three years (2005, 2010, 2014) 57 Figure 5-8 Average net primary production per land use in three years 58 Figure 6-1 Spatial and temporal resolution of both ecological processes and remote-sensing observation 69 ix List of Tables Table 1.1 Examples of regulation services Table 1.2 Ten proposed biodiversity variables in Skidmore et al (2015)’s framework Table 2.1 Ten proposed biodiversity variables by Skidmore et al (2015) 11 Table 2.2 Typical NDVI values for various cover types 13 Table 4.1 Radiometric range of bands and resolution for the ETM+ sensors 35 Table 4.2 Data preparation for this study 37 Table 4.3 ESUN value for Landsat sensor 39 Table 4.4 List of formulas and methods used in the paper 40 Table 4.5 Ordinal severity levels and example range of dNBR (scaled by 103) 43 Table 4.6 Illustration of approach used in ecosystem health assessment 44 Table 4.7 Illustration of weighted scoring approach for the ecosystem of vigor 45 Table 5.1 Leaf area index per land use in three time steps 50 Table 5.2 Coefficients of the polynomial relationship for Mo between T* and fc 51 Table 5.3 Soil moisture per land use in three time steps 54 Table 5.4 Net primary production per land use in three years 59 Table 5.5 Ecosystem health assessment of LAI in 2010 and 2015 61 Table 5.6 Ecosystem health assessment of soil moisture in 2010 and 2015 62 Table 5.7 Ecosystem health assessment of NPP in 2010 and 2014 63 Table 5.8 Weighted scoring for the vigor of ecosystem health in 2010 and 2015 64 Table 6.1 A combination between indicators of ecosystem health and 71 xi Society of America Retrieved March 19, 2016, from http://vzj.geoscienceworld.org/content/7/1/358.full Rocchini, D., Foody, G M., Nagendra, H., Ricotta, C., Anand, M., He, K S., Amici, V., et al (2013) Uncertainty in ecosystem mapping by remote sensing Computers and Geosciences, 50, 128–135 Roerink, G J., Menenti, M., & Verhoef, W (2000) Reconstructing cloudfree NDVI composites using Fourier analysis of time series International Journal of Remote Sensing, 21(9), 1911–1917 Retrieved from http://www.tandfonline.com/doi/abs/10.1080/014311600209814 Rouse, J W J., Haas, R H., Schell, J A., & Deering, D W (1974) Monitoring Vegetation Systems in the Great Plains with ERTS, Retrieved December 9, 2015, from https://www.researchgate.net/publication/246866091_Monitoring_Vegetation_Systems_i n_the_Great_Plains_with_ERTS Roy, D P., Boschetti, L., & Trigg, S N (2006) Remote sensing of fire severity: assessing the performance of the normalized burn ratio IEEE Geoscience and Remote Sensing Letters, 3(1), 112–116 Retrieved from http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1576701 Running, S W., Loveland, T R., Pierce, L L., Nemani, R R., & Hunt Jr., E R (1995) A remote sensing based vegetation classification logic for global land cover analysis Remote Sensing of Environment [REMOTE SENS ENVIRON.], 51(January 1994), 39–48 Running, S W., & Zhao, M (2015) User’s Guide Daily GPP and Annual NPP (MOD17A2/A3) Products NASA Earth Observing System MODIS Land Algorithm Schepers, L., Haest, B., Veraverbeke, S., Spanhove, T., Borre, J Vanden, & Goossens, R (2014) Burned area detection and burn severity assessment of a heathland fire in belgium using airborne imaging spectroscopy (APEX) Remote Sensing, 6(3), 1803–1826 Secretariat of the Conservation on Biological Diversity (2010) Aichi Biodiversity Targets Retrieved October 8, 2015, from https://www.cbd.int/sp/targets/ Skidmore, A K., Pettorelli, N., Coops, N C., Geller, G N., Hansen, M., Lucas, R., Mucher, C References 90 A., et al (2015) Agree on biodiveristy metrics to track from space Nature, 523, 403 – 405 Smith, L C (1997) Satellite remote sensing of river inundation area, stage, and discharge: A review Hydrological processes, 11(10), 1427–1439 Sobrino, J a., Jiménez-Muñoz, J C., & Paolini, L (2004) Land surface temperature retrieval from LANDSAT TM Remote Sensing of Environment, 90(4), 434–440 Sonnentag, O., Chen, J M., Roberts, D A., Talbot, J., Halligan, K Q., & Govind, A (2007) Mapping tree and shrub leaf area indices in an ombrotrophic peatland through multiple endmember spectral unmixing Remote Sensing of Environment, 109(3), 342–360 Spellman, F R (2015) Handbook of Environmental Engineering CRC Press Retrieved February 9, 2016, from https://books.google.com/books?id=dh6vCgAAQBAJ&pgis=1 Stenberg, P., Rautiainen, M., Manninen, T., Voipio, P., & Smolander, H (2004) Reduced simple ratio better than NDVI for estimating LAI in Finnish pine and spruce stands Silva Fennica, 38(December 2003), 3–14 Stork, N., & Samways, M (1995) Inventorying and Monitoring In R Watson (Ed.), Global Biodiversity Assessment Cambrigde University Press Strand, H., Hoft, R., Strittholt, J., Horning, N., Miles, L., Fosnight, E., & Turner, W (2007) Sourcebook on remote sensing and biodiversity indicators CBD Technical Series 32 Townshend, J., Justice, C., Li, W., Gurney, C., & McManus, J (1991) Global land cover classification by remote sensing: present capabilities and future possibilities Remote Sensing of Environment, 35(2-3), 243–255 Retrieved from http://www.sciencedirect.com/science/article/pii/003442579190016Y Turner, D P., Cohen, W B., Kennedy, R E., Fassnacht, K S., & Briggs, J M (1999) Relationships between leaf area index and Landsat TM spectral vegetation indices across three temperate zone sites Remote Sensing of Environment, 70(1), 52–68 USFS (2009) Operational Satellite-based remote sensing fire support program USDA Forest Service Remote Sensing Applications Center Retrieved March 3, 2016, from http://www.mtbs.gov/ References 91 USGS (2012) NLCD 92 Land cover class Definitions Retrieved October 8, 2015, from http://landcover.usgs.gov/classes.php#top Valor, E., & Caselles, V (1996) Mapping land surface emissivity from NDVI: Application to European, African, and South American areas Remote Sensing of Environment, 57(3), 167–184 Retrieved from http://www.sciencedirect.com/science/article/pii/0034425796000399 Vietnam Statistic Head Office (2013) Administrative Unit, Land and Climate Ha Noi van der Kwast, J (2016) Python scripts for Zonal Statistic and Scatter Plot Wegmüller, U., & Werner, C (1997) Retrieval of vegetation parameters with SAR interferometry IEEE Transactions on Geoscience and Remote Sensing, 35(1), 18–24 Whittaker, R H (1970) Communities and ecosystems London: MacMillan Publishing Co., Inc World Meteorological Organization (2013) Measurement of soil moisture Yang, X., Wu, J., Shi, P., & Yan, F (2008) Modified triangle method to estimate soil moisture status with MODerate resolution Imaging Spectroradiometer (MODIS) products Proc Int Archives Photogramm., Remote Sensing and Spatial Information Sciences, XXXVII(B8), 555–560 Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.184.3316&rep=rep1&type=pd f Zhang, J., Wang, Y., & Li, Y (2006) A C++ program for retrieving land surface temperature from the data of Landsat TM/ETM+ band6 Computers & Geosciences, 32(10), 1796– 1805 Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S0098300406000793 Zhang, X., Friedl, M a., Schaaf, C B., Strahler, A H., Hodges, J C F., Gao, F., Reed, B C., et al (2003) Monitoring vegetation phenology using MODIS Remote Sensing of Environment, 84(3), 471–475 Zhang, Y (2004) Understanding Image Fusion Photogrammetric Engineering & Remote Sensing, (June), 657–661 References 92 Zhao, M., Heinsch, F A., Nemani, R R., & Running, S W (2005) Improvements of the MODIS terrestrial gross and net primary production global data set Remote Sensing of Environment, 95, 164–176 Zheng, G., & Moskal, L M (2009) Retrieving Leaf Area Index (LAI) Using Remote Sensing: Theories, Methods and Sensors Sensors, 9(4), 2719–2745 Retrieved from http://www.mdpi.com/1424-8220/9/4/2719/ Zheng, N., Tachikawa, Y., & Takara, K (2008) A Distributed Flood Inundation Model Integrating With Rainfall-Runoff Processes Using Gis and Remote Sensing Data The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 1513–1518 Zhu, Y., Yao, X., Tian, Y., Liu, X., & Cao, W (2008) Analysis of common canopy vegetation indices for indicating leaf nitrogen accumulations in wheat and rice International Journal of Applied Earth Observation and Geoinformation, 10(1), 1–10 Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S0303243407000165 Zimdaht, R L (1980) Weed-crop competition: a review Corvallis, USA: International Plant Protection Center USGS Products Acknowledgement: Data available from the U.S Geological Survey See USGS Visual Identity System Guidance for further details Questions concerning the use or redistribution of USGS data should be directed to: ask@usgs.gov or 1-888-ASK-USGS (1-888-275-8747) NASA Land Processes Distributed Active Archive Centre (LP DAAC) Products Cover picture source: http://ngheanprovince.info/2015/08/13/travel-and-explore-pu-matnational-park-in-nghe-an/ References 93 Appendices Conversion DNs to TOA brightness temperature Digital numbers (DNs) are expressed as integer values to facilitate computation and transmission and to scale brightnesses for convenient display without physical units DNs cannot serve as input for models of physical processes in agriculture, forestry or hydrology because there is no defined relationship to other images or to features on the ground (Campbell & Wynne, 2011) As a consequence, radiances conversion from DNs should be carried out to form an essential transformation so as to prepare remotely sensed imagery for future analyses For a given sensor, spectral channel, and DN, the corresponding radiance value (L) can be calculated as 𝐿 = [(𝐿𝑚𝑎𝑥 − 𝐿𝑚𝑖𝑛 )/(𝑄𝑐𝑎𝑙𝑚𝑎𝑥 − 𝑄𝑐𝑎𝑙𝑚𝑖𝑛 )] × (𝑄𝑐𝑎𝑙 − 𝑄𝑐𝑎𝑙𝑚𝑖𝑛 ) + 𝐿𝑚𝑖𝑛 where Lλ is the spectral radiance at the sensors aperture [W/m2 sr µm]; Qcal is the quantized calibrated pixel value (DN); Qcalmin is the minimum quantized calibrated pixel value corresponding to Lminλ (DN); Qcalmax is the maximum quantized calibrated pixel value corresponding to Lmaxλ (DN); Lmin is the spectral at-sensor radiance that is scaled to Qcalmin [W/m2 sr µm]; and Lmax is the spectral at-sensor radiance that is scaled to Qcalmax [W/m2 sr µm] Calibration for ETM+ sensor has been presented by Chander et al (2009) or 𝐿𝜆 = 𝐺𝑟𝑒𝑠𝑐𝑎𝑙𝑒 × 𝑄𝑐𝑎𝑙 + 𝐵𝑟𝑒𝑠𝑐𝑎𝑙𝑒 Where 𝐺𝑟𝑒𝑠𝑐𝑎𝑙𝑒 = 𝐿𝑀𝐴𝑋𝜆 − 𝐿𝑀𝐼𝑁𝜆 𝑄𝑐𝑎𝑙𝑚𝑎𝑥 − 𝑄𝑐𝑎𝑙𝑚𝑖𝑛 𝐵𝑟𝑒𝑠𝑐𝑎𝑙𝑒 = 𝐿𝑀𝐼𝑁𝜆 − ( Appendices 𝐿𝑀𝐴𝑋𝜆 − 𝐿𝑀𝐼𝑁𝜆 )𝑄 𝑄𝑐𝑎𝑙𝑚𝑎𝑥 − 𝑄𝑐𝑎𝑙𝑚𝑖𝑛 𝑐𝑎𝑙𝑚𝑖𝑛 95 where Grescale is band-specific rescaling gain factor [(W/m2 sr µm)/DN]; Brescale is the band-specific rescaling bias factor [W/ (m2 sr µm)] As reported in Chander et al (2009)’s work, Grescale and Brescale are 0.037205 and 3.16 respectively (for high gain band 6) Conversion to at-sensor brightness temperature The thermal band data (band on ETM+) can be converted from at-sensor spectral radiance to effective at-sensor brightness temperature The at-sensor brightness temperature assumes that the Earth’s surface is a black body and includes atmospheric effects (absorption and emission along path) The conversion formula from the at-sensor’s spectral radiance to at-sensor brightness temperature is: 𝑇= 𝐾2 𝐾 ln( 𝐿1 + 1) 𝜆 where T is the at-satellite brightness temperature (K); Lλ is TOA spectral radiance [W/m2 sr μm)]; K1 is band-specific thermal conversion constant from the metadata [W/m2 sr μm)]; K2 is band-specific thermal conversion constant from the metadata (K) K1 [W/m2 sr μm)] and K2 (K) constant for Landsat sensor are 666.09 and 1282.71, respectively (NASA, 2011) Appendices 96 Python Scripts (van der Kwast, 2016) Python Script to executive Zonal Statistic # This script can be used to calculate zonal statistics # Usage: python zonalstats.py # Input (argument 1): land-use map in PCRaster format, containing the zones # Input (argument 2): continuous map in PCRaster format for which the stats will be calculated # Output (argument 3): table with zonal statistics from pcraster import * #import the PCRaster library import sys, os # library used to read arguments from the command line import csv # library needed to write the csv table setglobaloption("unitcell") #used to change global setting for no of cells #function to calculate zonal standard deviation def areastd(continuousmap,discretemap,average,numberofcells): variance = areatotal((sqr(continuousmap - average))/numberofcells,discretemap) std = sqrt(variance) return std def samplefrommap(luclass,samplelocation,average,minimum,maximum,std,numcells): luclass = str(luclass) zonalAverageAtSample = ifthen(defined(samplelocation),zonalAverage) zonalAverageAtSample = mapmaximum(zonalAverageAtSample) zonalAverageAtSample = cellvalue(zonalAverageAtSample,0,0) zonalAverageAtSample = str(zonalAverageAtSample[0]) zonalMinimumAtSample = ifthen(defined(samplelocation),zonalMinimum) zonalMinimumAtSample = mapmaximum(zonalMinimumAtSample) zonalMinimumAtSample = cellvalue(zonalMinimumAtSample,0,0) zonalMinimumAtSample = str(zonalMinimumAtSample[0]) Appendices 97 zonalMaximumAtSample = ifthen(defined(samplelocation),zonalMaximum) zonalMaximumAtSample = mapmaximum(zonalMaximumAtSample) zonalMaximumAtSample = cellvalue(zonalMaximumAtSample,0,0) zonalMaximumAtSample = str(zonalMaximumAtSample[0]) zonalNumberOfCellsAtSample = ifthen(defined(samplelocation),zonalNumberOfCells) zonalNumberOfCellsAtSample = mapmaximum(zonalNumberOfCellsAtSample) zonalNumberOfCellsAtSample = cellvalue(zonalNumberOfCellsAtSample,0,0) zonalNumberOfCellsAtSample = str(int(zonalNumberOfCellsAtSample[0])) zonalStdAtSample = ifthen(defined(samplelocation),zonalStd) zonalStdAtSample = mapmaximum(zonalStdAtSample) zonalStdAtSample = cellvalue(zonalStdAtSample,0,0) zonalStdAtSample = str(zonalStdAtSample[0]) Data=([luclass,zonalAverageAtSample,zonalMinimumAtSample,zonalMaximumAtSample,z onalStdAtSample,zonalNumberOfCellsAtSample]) Write.writerow(Data) #print to screen what is executed print "Calculating zonal statistics for " + str(sys.argv[2]) + " using the classes of " + str(sys.argv[1]) #read inputs from command line arguments LandUse = readmap(sys.argv[1]) ContinuousMap = readmap(sys.argv[2]) #read output filename, write header OutputTable = open(sys.argv[3],"w") Write = csv.writer(OutputTable) Header = (["landuse","average","min","max","std","N"]) Write.writerow(Header) Appendices 98 #calculate zonal statistics print "Calculating zonal average " zonalAverage = areaaverage(ContinuousMap,LandUse) print "Calculating zonal minimum " zonalMinimum = areaminimum(ContinuousMap,LandUse) print "Calculating zonal maximum " zonalMaximum = areamaximum(ContinuousMap,LandUse) print "Calculating zonal standard deviation " zonalNumberOfCells = areaarea(LandUse) zonalStd = areastd(ContinuousMap,LandUse,zonalAverage,zonalNumberOfCells) #preprocess data for table nrClasses = mapmaximum(scalar(LandUse)) #determine how many classes in map nrClasses = cellvalue(nrClasses,0,0) #convert the result from map to list nrClasses = int(nrClasses[0]) #convert the result from list to integer value SampleMap = ifthen(pcreq(LandUse,1000), LandUse) #create empty map for i in range (1, nrClasses + 1): #loop over classes Random = ifthen(pcreq(LandUse,i), + uniform(boolean(LandUse))) #create uniform distribution in class RandomOrder = order(Random) #order to get unique integer values Sample = ifthen(pcrle(RandomOrder,1), LandUse) #create map with selected pixel SampleMap = ifthen(defined(LandUse), cover(SampleMap, Sample)) #add selected pixel to other selected pixels samplefrommap(i,Sample,zonalAverage,zonalMinimum,zonalMaximum,zonalStd,zonalNum berOfCells) Percentage = (i / float(nrClasses)) * 100.0 print "Writing table:", int(Percentage), "% completed\r", print OutputTable.close() Appendices 99 Python Script to executive Scatter Plot # This script makes a scatter plot of a comma separated file # syntax: python plotscatter.py import csv import matplotlib.pyplot as plt import sys import numpy as np import pandas as pd def getColumn(filename, column): results = csv.reader(open(filename)) return [result[column] for result in results] def regression(xvalues,yvalues): dataset = pd.DataFrame.from_csv(inputFile, sep= ',', header=None) regressionCoefficients = np.polyfit(dataset[x],dataset[y],1) p = np.poly1d(regressionCoefficients) yhat = p(dataset[x]) ybar = np.sum(dataset[y])/len(dataset[y]) ssreg = np.sum((yhat-ybar)**2) sstot = np.sum((dataset[y]-ybar)**2) Rsq = ssreg/sstot print "Regression equation:" print "y =", regressionCoefficients[0], "x +", regressionCoefficients[1] print "R2 =", Rsq return regressionCoefficients inputFile = str(sys.argv[1]) xColumn = int(sys.argv[2]) yColumn = int(sys.argv[3]) x = xColumn - y = yColumn - Appendices 100 print "Calculating scatterplot " + sys.argv[4] print "Input CSV file: " + inputFile print "X-axis column: " + str(xColumn) + ", " + sys.argv[5] print "Y-axis column: " + str(yColumn) + ", " + sys.argv[6] valueXaxis = getColumn(inputFile,x) # first column = > valueYaxis = getColumn(inputFile,y) regressionResults = regression(x,y) plt.figure(str(sys.argv[4])) plt.title(str(sys.argv[4])) plt.xlabel(str(sys.argv[5])) plt.ylabel(str(sys.argv[6])) plt.scatter(valueXaxis,valueYaxis) axes = plt.gca() a, b = regressionResults X_plot = np.linspace(axes.get_xlim()[0],axes.get_xlim()[1],100) plt.plot(X_plot, a*X_plot + b, '-') plt.show() Appendices 101 Net primary production and World Net Primary productivity for major ecosystems Source: Adapted from Whittaker (1970) Net primary production per unit area (dry g/m2/year) World Net primary production Area (106 km2) Normal range Mean (109 dry tons/year) Lake and stream 100 – 1500 500 1.0 Swamp and march 800 – 4000 2000 4.0 Tropical forest 20 1000 – 5000 2000 40.0 Temperate forest 18 600 – 2500 1300 23.4 Boreal forest 12 400 – 2000 800 9.6 Woodland and shrub land 200 – 1200 600 4.2 Savana 45 200 – 2000 700 10.4 Temperate grassland 150 – 1500 500 4.5 Tundra and alpine 10 – 400 140 1.1 Desert scrub 18 10 – 250 70 1.3 Extreme desert, rock, ice 24 – 10 0.07 Agricultural land 14 100 – 4000 650 9.1 Total land 149 730 109.0 Open ocean 332 – 400 125 41.5 Continental shelf 27 200 – 600 350 9.5 Attached algea, estuaries 500 - 4000 2000 4.0 Total ocean 361 155 55.0 Total for Earth 510 320 164.0 Appendices 103 MODIS 17 The U.S National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) currently “produces a regular global estimate of daily gross primary productivity (GPP) and annual net primary production (NPP) of the entire terrestrial earth surface at 1-km spatial resolution, 110 million cells, each having GPP and NPP computed individually” (Zhao et al., 2005) The derivation of a satellite estimate of terrestrial NPP has three theoretical components: (1) the idea that plant NPP is directly related to absorbed solar energy, (2) the theory that a connection exists between absorbed solar energy and satellite-derived spectral indices of vegetation, and (3) the assumption that there is biophysical reason why the actual conversion efficiency of absorbed solar energy may be reduced below the theoretical potential value How are GPP and NPP calculated? GPP is the initial daily total of photosynthesis, and daily net photosynthesis (PSNnet) subtracts leaf and fine-root respiration over a 24-hour day NPP is the annual sum of daily net PSN minus the cost of growth and maintenance of living cells in permanent woody tissue ## To estimate GPP/NPP, the main data inputs to the MOD17 algorithm include: - Fraction of Photosynthetically Active Radiation (FPAR) and Leaf Area Index (LAI) from the MODIS MOD15LAI/FPAR data product - Temperature, incoming solar radiation, and vapour pressure deficit derived from a meteorology dataset Meteorology dataset used by a various version of the MOD17 algorithms include products from the NASA Global Modelling and Assimilation Office (GMAO) and the NCEP/NCAR Reanalysis II - Land cover classification from the MODIS MCD12Q1 data product - A Biome Parameter Look-Up Table (BPLUT) containing the values of ɛmax for different vegetation types and other biome-specific physiological parameters for respiration calculations The different vegetation types (obtained from the land cover type classification) include: evergreen needleleaf forest, evergreen broadleaf forest, deciduous needleleaf forest, deciduous broadleaf forest, mixed forests, closed shrublands, open shrublands, woody savannas, savannas, grasslands, and croplands Appendices 104 Flowcharts showing the logic behind the MOD17 Algorithm in calculating both 8-day average GPP and annual NPP (Source: Running & Zhao (2015)) Biophysical variability of ɛ The photosynthetically active radiation (PAR) conversion efficiency ɛ includes two principle sources First, with any vegetation, some photosynthesis is immediately used for maintenance respiration The second source of variability ɛ is attributed to suboptimal climatic condition There are five parameters used to calculate GPP: ɛ, TMINmax, TMINmin, VPDmax and VPDmin Values of TMIN and VPD are obtain from GMAO/NASA dataset, while the value of ɛmax is obtained from Biome Properties Look-Up Table (BPLUT) The resulting radiation use efficiency coefficient ɛ is combined with estimates of APAR to calculate GPP (kg C day-1) as GPP = ɛ * APAR Appendices (1) 105 Where APAR = IPAR * FPAR, IPAR (PAR incident on the vegetative surface) must be estimated from incident shortwave radiation (SWRad, provided in the GMAO/NASA dataset) as IPAR = SWRad * 0.45 (2) While GPP (Equation 1) is calculated on a daily basis, 8-day summations of GPP are created and these summations are available to the public The summations are named for the first day included in the 8-day period Each summation consists of consecutive days of data To obtain an estimate of daily GPP for this 8-day period, it is necessary to divide the value obtained during a data download by eight (in the first 45 values/year) and by five (or six in a leap year) for final period ** Maintenance respiration costs (MR) for leaves and fine roots are summarized in the centre flowchart of Figure X and are also calculated on a daily basis ** Annual net primary production (NPP yearly) and growth respiration To calculate NPP, MOD17 also estimates daily leaf and fine root maintenance respiration (R-ir), annual growth respiration (Rg), and annual maintenance respiration of live cells in woody tissue (Rm) Rg is the energy cost for constructing organic compounds fixed by photosynthesis, and it is empirically parameterized as 25% as NPP To improve the algorithm, LAImax dependent Rg was replaced with Rg = 0.25 * NPP and annual MODIS NPP can be computed as NPP = GPP – Rm – 0.25*NPP (3) where Rm is annual plant maintenance respiration, and therefore, NPP = 0.8 * (GPP – Rm) when GPP – Rm ≥ 0, and NPP = when GPP – Rm < Appendices (4) 106