GEE-TED: A tsetse ecological distribution model for Google Earth Engine

GEE-TED: A tsetse ecological distribution model for Google Earth Engine Associated publication forthcoming: Fox, L., Peter, B. G., Frake, A. N., and Messina, J. P. (Forthcoming). A Bayesian maximum entropy model for predicting tsetse ecological distributions. Journal TBD. Description GEE-TED is a Google Earth Engine (GEE; Gorelick et al. 2017) adaptation of a tsetse ecological distribution (TED) model developed by DeVisser et al. (2010), which was designed for use in ESRI's ArcGIS. TED uses time-series climate and land-use/land-cover (LULC) data to predict the probability of tsetse presence across space based on species habitat preferences (in this case Glossina Morsitans). Model parameterization includes (1) day and night temperatures (MODIS Land Surface Temperature; MOD11A2), (2) available moisture/humidity using a vegetation index as a proxry (MODIS NDVI; MOD13Q1), (3) LULC (MODIS Land Cover Type 1; MCD12Q1), (4) year selections, and (5) fly movement rate (meters/16-days). TED has also been used as a basis for the development of an agent-based model by Lin et al. (2015) and in a cost-benefit analysis of tsetse control in Tanzania by Yang et al. (2017). Parameterization in Fox et al. (Forthcoming): Suitable LULC types and climate thresholds used here are specific to Glossina Morsitans in Kenya and are based on the parameterization selections in DeVisser et al. (2010) and DeVisser and Messina (2009). Suitable temperatures range from 17–40°C during the day and 10–40°C at night and available moisture is characterized as NDVI > 0.39. Suitable LULC comprises predominantly woody vegetation; a complete list of suitable categories is available in DeVisser and Messina (2009). In the Fox et al. (Forthcoming) publication, two versions of MCD12Q1 were used to assess suitable LULC types: Versions 051 and 006. The GeoTIFF supplied in this dataset entry (GEE-TED_Kenya_2016-2017.tif) uses the aforementioned parameters to show the probable tsetse distribution across Kenya for the years 2016-2017. A static graphic of this GEE-TED output is shown below and an interactive version can be viewed at: https://cartoscience.users.earthengine.app/view/gee-ted. Figure associated with Fox et al. (Forthcoming) GEE code The code supplied below is generalizable across geographies and species; however, it is highly recommended that parameterization is given considerable attention to produce reliable results. Note that output visualization on-the-fly will take some time and it is recommended that results be exported as an asset within GEE or exported as a GeoTIFF. Note: Since completing the Fox et al. (Forthcoming) manuscript, GEE has removed Version 051 per NASA's deprecation of the product. The current release of GEE-TED now uses only MCD12Q1 Version 006; however, alternative LULC data selections can be used with minimal modification to the code. // Input options var tempMin = 10 // Temperature thresholds in degrees Celsius var tempMax = 40 var ndviMin = 0.39 // NDVI thresholds; proxy for available moisture/humidity var ndviMax = 1 var movement = 500 // Fly movement rate in meters/16-days var startYear = 2008 // The first 2 years will be used for model initialization var endYear = 2019 // Computed probability is based on startYear+2 to endYear var country = 'KE' // Country codes - https://en.wikipedia.org/wiki/List_of_FIPS_country_codes var crs = 'EPSG:32737' // See https://epsg.io/ for appropriate country UTM zone var rescale = 250 // Output spatial resolution var labelSuffix = '02052020' // For file export labeling only //[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17] MODIS/006/MCD12Q1 var lulcOptions006 = [1,1,1,1,1,1,1,1,1, 0, 1, 0, 0, 0, 0, 0, 0] // 1 = suitable 0 = unsuitable // No more input required ------------------------------ // var region = ee.FeatureCollection(\"USDOS/LSIB_SIMPLE/2017\") .filterMetadata('country_co', 'equals', country) // Input parameter modifications var tempMinMod = (tempMin+273.15)/0.02 var tempMaxMod = (tempMax+273.15)/0.02 var ndviMinMod = ndviMin*10000 var ndviMaxMod = ndviMax*10000 var ndviResolution = 250 var movementRate = movement+(ndviResolution/2) // Loading image collections var lst = ee.ImageCollection('MODIS/006/MOD11A2').select('LST_Day_1km', 'LST_Night_1km') .filter(ee.Filter.calendarRange(startYear,endYear,'year')) var ndvi = ee.ImageCollection('MODIS/006/MOD13Q1').select('NDVI') .filter(ee.Filter.calendarRange(startYear,endYear,'year')) var lulc006 = ee.ImageCollection('MODIS/006/MCD12Q1').select('LC_Type1') // Lulc mode and boolean reclassification var lulcMask = lulc006.mode().remap([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17],lulcOptions006) .eq(1).rename('remapped').clip(region) // Merge NDVI and LST image collections var combined = ndvi.combine(lst, true) var combinedList = combined.toList(10000) // Boolean reclassifications (suitable/unsuitable) for day/night temperatures and ndvi var con = function(image) { var ndviExp = image.expression(\"(b('NDVI') > \"+ndviMaxMod+\") ? 0\" + \":(b('NDVI') > \"+ndviMinMod+\") ? 1\" + \":(b('NDVI') < \"+ndviMinMod+\") ? 0\" + \":1\").rename('ndvi_suit') var lstDayExp = image.expression(\"(b('LST_Day_1km') > \"+tempMaxMod+\") ? 0\" + \":(b('LST_Day_1km') > \"+tempMinMod+\") ? 1\" + \":(b('LST_Day_1km') < \"+tempMinMod+\") ? 0\" + \":1\").rename('day_suit') var lstNightExp = image.expression(\"(b('LST_Night_1km') > \"+tempMaxMod+\") ? 0\" + \":(b('LST_Night_1km') > \"+tempMinMod+\") ? 1\" + \":(b('LST_Night_1km') < \"+tempMinMod+\") ? 0\" + \":1\").rename('night_suit') var add = image.addBands(ndviExp).addBands(lstDayExp).addBands(lstNightExp) var multiply = add.select('ndvi_suit').multiply(add.select('day_suit')).multiply(add.select('night_suit')) return add.addBands(multiply.rename('suit')).select('suit') } var conList = combined.map(con).toList(10000) var finish = conList.size().subtract(1).getInfo() // Fly movement rate model var iterateList = conList var kernel = ee.Kernel.square({radius: movementRate, units: 'meters'}) var fill = ee.Image(conList.get(0)).multiply(0) var expansion = fill.add(1) for (var range = 0; range <= finish; range = range + 1) { var img = ee.Image(iterateList.get(range)).select('suit').gt(0).clip(region) var multi = img.multiply(expansion).set('num',range) var mask = multi.eq(1) var masked = multi.updateMask(mask) var expand = masked.focal_max({kernel: kernel}) var expansion = fill.where(expand,1) var updateList = iterateList.add(multi) var iterateList = updateList } // Filter out first two years and compute tsetse probability var filtered = ee.ImageCollection(updateList).sort('num',false).limit(updateList.size().divide(2).subtract(46)) var filteredProb = filtered.sum().divide(filtered.size()).multiply(100).updateMask(lulcMask) var probMask = filteredProb.neq(0) var filteredProbMasked = filteredProb.updateMask(probMask).rename('probability') var startMod = startYear+2 var naming = 'GEE-TED_'+country+'-'+startMod+'-'+endYear+'_'+labelSuffix Map.addLayer(lulcMask, {min:0, max:1}, 'MCD12Q1 LULC mask', false) Map.addLayer(filteredProbMasked, {palette: ['FCFAE1','F6E497','BD8D46','B9121B','4C1B1B'], opacity: 0.8, min: 0, max: 100}, naming, false) Map.setOptions('HYBRID') Map.centerObject(region) // Export image to drive Export.image.toDrive({ image: filteredProbMasked, description: naming, fileNamePrefix: naming, maxPixels: 1e13, scale: rescale, crs: crs, region: region }) References • DeVisser, M.H. and Messina, J.P., 2009. Optimum land cover products for use in a Glossina-morsitans habitat model of Kenya. International Journal of Health Geographics, 8(1), pp.1-20. • DeVisser, M.H., Messina, J.P., Moore, N.J., Lusch, D.P. and Maitima, J., 2010. A dynamic species distribution model of Glossina subgenus Morsitans: The identification of tsetse reservoirs and refugia. Ecosphere, 1(1), pp.1-21. • Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D. and Moore, R., 2017. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment, 202, pp.18-27. • Lin, S., DeVisser, M.H. and Messina, J.P., 2015. An agent-based model to simulate tsetse fly distribution and control techniques: a case study in Nguruman, Kenya. Ecological Modelling, 314, pp.80-89. • Yang, A., Messina, J.P., Grady, S.C. and White, R.A., 2017. Cost–Benefit Analysis of Tsetse Fly Control in Tanzania. Papers in Applied Geography, 3(2), pp.182-195.