def main(): try: import sklearn import joblib if sklearn.__version__ < "0.20": gs.fatal( "Package python3-scikit-learn 0.20 or newer is not installed") except ImportError: gs.fatal("Package python3-scikit-learn 0.20 or newer is not installed") # parser options group = options["group"] output = options["output"] model_load = options["load_model"] probability = flags["p"] prob_only = flags["z"] chunksize = int(options["chunksize"]) # remove @ from output in case overwriting result if "@" in output: output = output.split("@")[0] # check probabilities=True if prob_only=True if prob_only is True and probability is False: gs.fatal("Need to set probabilities=True if prob_only=True") # reload fitted model and training data estimator, y, class_labels = joblib.load(model_load) # define RasterStack stack = RasterStack(group=group) # perform raster prediction region = Region() row_incr = math.ceil(chunksize / region.cols) # do not read by increments if increment > n_rows if row_incr >= region.rows: row_incr = None # prediction if prob_only is False: gs.message("Predicting classification/regression raster...") stack.predict( estimator=estimator, output=output, height=row_incr, overwrite=gs.overwrite(), ) if probability is True: gs.message("Predicting class probabilities...") stack.predict_proba( estimator=estimator, output=output, class_labels=np.unique(y), overwrite=gs.overwrite(), height=row_incr, ) # assign categories for classification map if class_labels is not None and prob_only is False: rules = [] for val, lab in class_labels.items(): rules.append(",".join([str(val), lab])) rules = "\n".join(rules) rules_file = string_to_rules(rules) r.category(map=output, rules=rules_file, separator="comma")
def main(): elevation = options['elevation'] slope = options['slope'] flat_thres = float(options['flat_thres']) curv_thres = float(options['curv_thres']) filter_size = int(options['filter_size']) counting_size = int(options['counting_size']) nclasses = int(options['classes']) texture = options['texture'] convexity = options['convexity'] concavity = options['concavity'] features = options['features'] # remove mapset from output name in case of overwriting existing map texture = texture.split('@')[0] convexity = convexity.split('@')[0] concavity = concavity.split('@')[0] features = features.split('@')[0] # store current region settings global current_reg current_reg = parse_key_val(g.region(flags='pg', stdout_=PIPE).outputs.stdout) del current_reg['projection'] del current_reg['zone'] del current_reg['cells'] # check for existing mask and backup if found global mask_test mask_test = gs.list_grouped( type='rast', pattern='MASK')[gs.gisenv()['MAPSET']] if mask_test: global original_mask original_mask = temp_map('tmp_original_mask') g.copy(raster=['MASK', original_mask]) # error checking if flat_thres < 0: gs.fatal('Parameter thres cannot be negative') if filter_size % 2 == 0 or counting_size % 2 == 0: gs.fatal( 'Filter or counting windows require an odd-numbered window size') if filter_size >= counting_size: gs.fatal( 'Filter size needs to be smaller than the counting window size') if features != '' and slope == '': gs.fatal('Need to supply a slope raster in order to produce the terrain classification') # Terrain Surface Texture ------------------------------------------------- # smooth the dem gs.message("Calculating terrain surface texture...") gs.message( "1. Smoothing input DEM with a {n}x{n} median filter...".format( n=filter_size)) filtered_dem = temp_map('tmp_filtered_dem') gs.run_command("r.neighbors", input = elevation, method = "median", size = filter_size, output = filtered_dem, flags='c', quiet=True) # extract the pits and peaks based on the threshold pitpeaks = temp_map('tmp_pitpeaks') gs.message("2. Extracting pits and peaks with difference > thres...") r.mapcalc(expression='{x} = if ( abs({dem}-{median})>{thres}, 1, 0)'.format( x=pitpeaks, dem=elevation, thres=flat_thres, median=filtered_dem), quiet=True) # calculate density of pits and peaks gs.message("3. Using resampling filter to create terrain texture...") window_radius = (counting_size-1)/2 y_radius = float(current_reg['ewres'])*window_radius x_radius = float(current_reg['nsres'])*window_radius resample = temp_map('tmp_density') r.resamp_filter(input=pitpeaks, output=resample, filter=['bartlett','gauss'], radius=[x_radius,y_radius], quiet=True) # convert to percentage gs.message("4. Converting to percentage...") r.mask(raster=elevation, overwrite=True, quiet=True) r.mapcalc(expression='{x} = float({y} * 100)'.format(x=texture, y=resample), quiet=True) r.mask(flags='r', quiet=True) r.colors(map=texture, color='haxby', quiet=True) # Terrain convexity/concavity --------------------------------------------- # surface curvature using lacplacian filter gs.message("Calculating terrain convexity and concavity...") gs.message("1. Calculating terrain curvature using laplacian filter...") # grow the map to remove border effects and run laplacian filter dem_grown = temp_map('tmp_elevation_grown') laplacian = temp_map('tmp_laplacian') g.region(n=float(current_reg['n']) + (float(current_reg['nsres']) * filter_size), s=float(current_reg['s']) - (float(current_reg['nsres']) * filter_size), w=float(current_reg['w']) - (float(current_reg['ewres']) * filter_size), e=float(current_reg['e']) + (float(current_reg['ewres']) * filter_size)) r.grow(input=elevation, output=dem_grown, radius=filter_size, quiet=True) r.mfilter( input=dem_grown, output=laplacian, filter=string_to_rules(laplacian_matrix(filter_size)), quiet=True) # extract convex and concave pixels gs.message("2. Extracting convexities and concavities...") convexities = temp_map('tmp_convexities') concavities = temp_map('tmp_concavities') r.mapcalc( expression='{x} = if({laplacian}>{thres}, 1, 0)'\ .format(x=convexities, laplacian=laplacian, thres=curv_thres), quiet=True) r.mapcalc( expression='{x} = if({laplacian}<-{thres}, 1, 0)'\ .format(x=concavities, laplacian=laplacian, thres=curv_thres), quiet=True) # calculate density of convexities and concavities gs.message("3. Using resampling filter to create surface convexity/concavity...") resample_convex = temp_map('tmp_convex') resample_concav = temp_map('tmp_concav') r.resamp_filter(input=convexities, output=resample_convex, filter=['bartlett','gauss'], radius=[x_radius,y_radius], quiet=True) r.resamp_filter(input=concavities, output=resample_concav, filter=['bartlett','gauss'], radius=[x_radius,y_radius], quiet=True) # convert to percentages gs.message("4. Converting to percentages...") g.region(**current_reg) r.mask(raster=elevation, overwrite=True, quiet=True) r.mapcalc(expression='{x} = float({y} * 100)'.format(x=convexity, y=resample_convex), quiet=True) r.mapcalc(expression='{x} = float({y} * 100)'.format(x=concavity, y=resample_concav), quiet=True) r.mask(flags='r', quiet=True) # set colors r.colors_stddev(map=convexity, quiet=True) r.colors_stddev(map=concavity, quiet=True) # Terrain classification Flowchart----------------------------------------- if features != '': gs.message("Performing terrain surface classification...") # level 1 produces classes 1 thru 8 # level 2 produces classes 5 thru 12 # level 3 produces classes 9 thru 16 if nclasses == 8: levels = 1 if nclasses == 12: levels = 2 if nclasses == 16: levels = 3 classif = [] for level in range(levels): # mask previous classes x:x+4 if level != 0: min_cla = (4*(level+1))-4 clf_msk = temp_map('tmp_clf_mask') rules = '1:{0}:1'.format(min_cla) r.recode( input=classif[level-1], output=clf_msk, rules=string_to_rules(rules), overwrite=True) r.mask(raster=clf_msk, flags='i', quiet=True, overwrite=True) # image statistics smean = r.univar( map=slope, flags='g', stdout_=PIPE).outputs.stdout.split(os.linesep) smean = [i for i in smean if i.startswith('mean=') is True][0].split('=')[1] cmean = r.univar( map=convexity, flags='g', stdout_=PIPE).outputs.stdout.split(os.linesep) cmean = [i for i in cmean if i.startswith('mean=') is True][0].split('=')[1] tmean = r.univar( map=texture, flags='g', stdout_=PIPE).outputs.stdout.split(os.linesep) tmean = [i for i in tmean if i.startswith('mean=') is True][0].split('=')[1] classif.append(temp_map('tmp_classes')) if level != 0: r.mask(flags='r', quiet=True) classification(level+1, slope, smean, texture, tmean, convexity, cmean, classif[level]) # combine decision trees merged = [] for level in range(0, levels): if level > 0: min_cla = (4*(level+1))-4 merged.append(temp_map('tmp_merged')) r.mapcalc( expression='{x} = if({a}>{min}, {b}, {a})'.format( x=merged[level], min=min_cla, a=merged[level-1], b=classif[level])) else: merged.append(classif[level]) g.rename(raster=[merged[-1], features], quiet=True) del TMP_RAST[-1] # Write metadata ---------------------------------------------------------- history = 'r.terrain.texture ' for key,val in options.iteritems(): history += key + '=' + str(val) + ' ' r.support(map=texture, title=texture, description='generated by r.terrain.texture', history=history) r.support(map=convexity, title=convexity, description='generated by r.terrain.texture', history=history) r.support(map=concavity, title=concavity, description='generated by r.terrain.texture', history=history) if features != '': r.support(map=features, title=features, description='generated by r.terrain.texture', history=history) # write color and category rules to tempfiles r.category( map=features, rules=string_to_rules(categories(nclasses)), separator='pipe') r.colors( map=features, rules=string_to_rules(colors(nclasses)), quiet=True) return 0
def compute_supply( base, recreation_spectrum, highest_spectrum, base_reclassification_rules, reclassified_base, reclassified_base_title, flow, flow_map_name, aggregation, ns_resolution, ew_resolution, print_only=False, flow_column_name=None, vector=None, supply_filename=None, use_filename=None, ): """ Algorithmic description of the "Contribution of Ecosysten Types" # FIXME ''' 1 B ← {0, .., m-1} : Set of aggregational boundaries 2 T ← {0, .., n-1} : Set of land cover types 3 WE ← 0 : Set of weighted extents 4 R ← 0 : Set of fractions 5 F ← 0 6 MASK ← HQR : High Quality Recreation 7 foreach {b} ⊆ B do : for each aggregational boundary 'b' 8 RB ← 0 9 foreach {t} ⊆ T do : for each Land Type 10 WEt ← Et * Wt : Weighted Extent = Extent(t) * Weight(t) 11 WE ← WE⋃{WEt} : Add to set of Weighted Extents 12 S ← ∑t∈WEt 13 foreach t ← T do 14 Rt ← WEt / ∑WE 15 R ← R⋃{Rt} 16 RB ← RB⋃{R} ''' # FIXME Parameters ---------- recreation_spectrum: Map scoring access to and quality of recreation highest_spectrum : Expected is a map of areas with highest recreational value (category 9 as per the report ... ) base : Base land types map for final zonal statistics. Specifically to ESTIMAP's recrceation mapping algorithm base_reclassification_rules : Reclassification rules for the input base map reclassified_base : Name for the reclassified base cover map reclassified_base_title : Title for the reclassified base map ecosystem_types : flow : Map of visits, derived from the mobility function, depicting the number of people living inside zones 0, 1, 2, 3. Used as a cover map for zonal statistics. flow_map_name : A name for the 'flow' map. This is required when the 'flow' input option is not defined by the user, yet some of the requested outputs required first the production of the 'flow' map. An example is the request for a supply table without requesting the 'flow' map itself. aggregation : ns_resolution : ew_resolution : statistics_filename : supply_filename : Name for CSV output file of the supply table use_filename : Name for CSV output file of the use table flow_column_name : Name for column to populate with 'flow' values vector : If 'vector' is given, a vector map of the 'flow' along with appropriate attributes will be produced. ? : Land cover class percentages in ROS9 (this is: relative percentage) output : Supply table (distribution of flow for each land cover class) Returns ------- This function produces a map to base the production of a supply table in form of CSV. Examples -------- """ # Inputs flow_in_base = flow + "_" + base base_scores = base + ".scores" # Define lists and dictionaries to hold intermediate data statistics_dictionary = {} weighted_extents = {} flows = [] # MASK areas of high quality recreation r.mask(raster=highest_spectrum, overwrite=True, quiet=True) # Reclassify land cover map to MAES ecosystem types r.reclass( input=base, rules=base_reclassification_rules, output=reclassified_base, quiet=True, ) # add to "remove_at_exit" after the reclassified maps! # Discard areas out of MASK copy_equation = EQUATION.format(result=reclassified_base, expression=reclassified_base) r.mapcalc(copy_equation, overwrite=True) # Count flow within each land cover category r.stats_zonal( base=base, flags="r", cover=flow_map_name, method="sum", output=flow_in_base, overwrite=True, quiet=True, ) # Set colors for "flow" map r.colors(map=flow_in_base, color=MOBILITY_COLORS, quiet=True) # Parse aggregation raster categories and labels categories = grass.parse_command("r.category", map=aggregation, delimiter="\t") for category in categories: # Intermediate names cells = highest_spectrum + ".cells" + "." + category remove_map_at_exit(cells) extent = highest_spectrum + ".extent" + "." + category remove_map_at_exit(extent) weighted = highest_spectrum + ".weighted" + "." + category remove_map_at_exit(weighted) fractions = base + ".fractions" + "." + category remove_map_at_exit(fractions) flow_category = "_flow_" + category flow = base + flow_category remove_map_at_exit(flow) flow_in_reclassified_base = reclassified_base + "_flow" flow_in_category = reclassified_base + flow_category flows.append(flow_in_category) # add to list for patching remove_map_at_exit(flow_in_category) # Output names msg = "Processing aggregation raster category: {r}" msg = msg.format(r=category) grass.debug(_(msg)) # g.message(_(msg)) # First, set region to extent of the aggregation map # and resolution to the one of the population map # Note the `-a` flag to g.region: ? # To safely modify the region: grass.use_temp_region() # FIXME g.region( raster=aggregation, nsres=ns_resolution, ewres=ew_resolution, flags="a", quiet=True, ) msg = "|! Computational resolution matched to {raster}" msg = msg.format(raster=aggregation) grass.debug(_(msg)) # Build MASK for current category & high quality recreation areas msg = "Setting category '{c}' of '{a}' as a MASK" grass.verbose(_(msg.format(c=category, a=aggregation))) masking = "if( {spectrum} == {highest_quality_category} && " masking += "{aggregation} == {category}, " masking += "1, null() )" masking = masking.format( spectrum=recreation_spectrum, highest_quality_category=HIGHEST_RECREATION_CATEGORY, aggregation=aggregation, category=category, ) masking_equation = EQUATION.format(result="MASK", expression=masking) grass.mapcalc(masking_equation, overwrite=True) # zoom to MASK g.region(zoom="MASK", nsres=ns_resolution, ewres=ew_resolution, quiet=True) # Count number of cells within each land category r.stats_zonal( flags="r", base=base, cover=highest_spectrum, method="count", output=cells, overwrite=True, quiet=True, ) cells_categories = grass.parse_command("r.category", map=cells, delimiter="\t") grass.debug(_("Cells: {c}".format(c=cells_categories))) # Build cell category and label rules for `r.category` cells_rules = "\n".join([ "{0}:{1}".format(key, value) for key, value in cells_categories.items() ]) # Discard areas out of MASK copy_equation = EQUATION.format(result=cells, expression=cells) r.mapcalc(copy_equation, overwrite=True) # Reassign cell category labels r.category(map=cells, rules="-", stdin=cells_rules, separator=":") # Compute extent of each land category extent_expression = "@{cells} * area()" extent_expression = extent_expression.format(cells=cells) extent_equation = EQUATION.format(result=extent, expression=extent_expression) r.mapcalc(extent_equation, overwrite=True) # Write extent figures as labels r.stats_zonal( flags="r", base=base, cover=extent, method="average", output=extent, overwrite=True, verbose=False, quiet=True, ) # Write land suitability scores as an ASCII file temporary_reclassified_base_map = temporary_filename( filename=reclassified_base) suitability_scores_as_labels = string_to_file( SUITABILITY_SCORES_LABELS, filename=temporary_reclassified_base_map) remove_files_at_exit(suitability_scores_as_labels) # Write scores as raster category labels r.reclass( input=base, output=base_scores, rules=suitability_scores_as_labels, overwrite=True, quiet=True, verbose=False, ) remove_map_at_exit(base_scores) # Compute weighted extents weighted_expression = "@{extent} * float(@{scores})" weighted_expression = weighted_expression.format(extent=extent, scores=base_scores) weighted_equation = EQUATION.format(result=weighted, expression=weighted_expression) r.mapcalc(weighted_equation, overwrite=True) # Write weighted extent figures as labels r.stats_zonal( flags="r", base=base, cover=weighted, method="average", output=weighted, overwrite=True, verbose=False, quiet=True, ) # Get weighted extents in a dictionary weighted_extents = grass.parse_command("r.category", map=weighted, delimiter="\t") # Compute the sum of all weighted extents and add to dictionary category_sum = sum([ float(x) if not math.isnan(float(x)) else 0 for x in weighted_extents.values() ]) weighted_extents["sum"] = category_sum # Create a map to hold fractions of each weighted extent to the sum # See also: # https://grasswiki.osgeo.org/wiki/LANDSAT#Hint:_Minimal_disk_space_copies r.reclass( input=base, output=fractions, rules="-", stdin="*=*", verbose=False, quiet=True, ) # Compute weighted fractions of land types fraction_category_label = { key: float(value) / weighted_extents["sum"] for (key, value) in weighted_extents.iteritems() if key is not "sum" } # Build fraction category and label rules for `r.category` fraction_rules = "\n".join([ "{0}:{1}".format(key, value) for key, value in fraction_category_label.items() ]) # Set rules r.category(map=fractions, rules="-", stdin=fraction_rules, separator=":") # Assert that sum of fractions is ~1 fraction_categories = grass.parse_command("r.category", map=fractions, delimiter="\t") fractions_sum = sum([ float(x) if not math.isnan(float(x)) else 0 for x in fraction_categories.values() ]) msg = "Fractions: {f}".format(f=fraction_categories) grass.debug(_(msg)) # g.message(_("Sum: {:.17g}".format(fractions_sum))) assert abs(fractions_sum - 1) < 1.0e-6, "Sum of fractions is != 1" # Compute flow flow_expression = "@{fractions} * @{flow}" flow_expression = flow_expression.format(fractions=fractions, flow=flow_in_base) flow_equation = EQUATION.format(result=flow, expression=flow_expression) r.mapcalc(flow_equation, overwrite=True) # Write flow figures as raster category labels r.stats_zonal( base=reclassified_base, flags="r", cover=flow, method="sum", output=flow_in_category, overwrite=True, verbose=False, quiet=True, ) # Parse flow categories and labels flow_categories = grass.parse_command("r.category", map=flow_in_category, delimiter="\t") grass.debug(_("Flow: {c}".format(c=flow_categories))) # Build flow category and label rules for `r.category` flow_rules = "\n".join([ "{0}:{1}".format(key, value) for key, value in flow_categories.items() ]) # Discard areas out of MASK # Check here again! # Output patch of all flow maps? copy_equation = EQUATION.format(result=flow_in_category, expression=flow_in_category) r.mapcalc(copy_equation, overwrite=True) # Reassign cell category labels r.category(map=flow_in_category, rules="-", stdin=flow_rules, separator=":") # Update title reclassified_base_title += " " + category r.support(flow_in_category, title=reclassified_base_title) # debugging # r.report( # flags='hn', # map=(flow_in_category), # units=('k','c','p'), # ) if print_only: r.stats( input=(flow_in_category), output="-", flags="nacpl", separator=COMMA, quiet=True, ) if not print_only: if flow_column_name: flow_column_prefix = flow_column_name + category else: flow_column_name = "flow" flow_column_prefix = flow_column_name + category # Produce vector map(s) if vector: # The following is wrong # update_vector(vector=vector, # raster=flow_in_category, # methods=METHODS, # column_prefix=flow_column_prefix) # What can be done? # Maybe update columns of an existing map from the columns of # the following vectorised raster map(s) # ? raster_to_vector(raster=flow_in_category, vector=flow_in_category, type="area") # get statistics dictionary = get_raster_statistics( map_one=aggregation, # reclassified_base map_two=flow_in_category, separator="|", flags="nlcap", ) # merge 'dictionary' with global 'statistics_dictionary' statistics_dictionary = merge_two_dictionaries( statistics_dictionary, dictionary) # It is important to remove the MASK! r.mask(flags="r", quiet=True) # FIXME # Add "reclassified_base" map to "remove_at_exit" here, so as to be after # all reclassified maps that derive from it # remove the map 'reclassified_base' # g.remove(flags='f', type='raster', name=reclassified_base, quiet=True) # remove_map_at_exit(reclassified_base) if not print_only: r.patch(flags="", input=flows, output=flow_in_reclassified_base, quiet=True) if vector: # Patch all flow vector maps in one v.patch( flags="e", input=flows, output=flow_in_reclassified_base, overwrite=True, quiet=True, ) # export to csv if supply_filename: supply_filename += CSV_EXTENSION nested_dictionary_to_csv(supply_filename, statistics_dictionary) if use_filename: use_filename += CSV_EXTENSION uses = compile_use_table(statistics_dictionary) dictionary_to_csv(use_filename, uses) # Maybe return list of flow maps? Requires unique flow map names return flows
def export_map(input_name, title, categories, colors, output_name, timestamp): """ Export a raster map by renaming the (temporary) raster map name 'input_name' to the requested output raster map name 'output_name'. This function is (mainly) used to export either of the intermediate recreation 'potential' or 'opportunity' maps. Parameters ---------- raster : Input raster map name title : Title for the output raster map categories : Categories and labels for the output raster map colors : Colors for the output raster map output_name : Output raster map name Returns ------- output_name : This function will return the requested 'output_name' Examples -------- .. """ finding = grass.find_file(name=input_name, element="cell") if not finding["file"]: grass.fatal("Raster map {name} not found".format( name=input_name)) # Maybe use 'finding'? # inform msg = "* Outputting '{raster}' map\n" msg = msg.format(raster=output_name) grass.verbose(_(msg)) # get categories and labels temporary_raster_categories_map = temporary_filename("categories_of_" + input_name) raster_category_labels = string_to_file( string=categories, filename=temporary_raster_categories_map) # add ascii file to removal list remove_files_at_exit(raster_category_labels) # apply categories and description r.category(map=input_name, rules=raster_category_labels, separator=":") # update meta and colors update_meta(input_name, title, timestamp) r.colors(map=input_name, rules="-", stdin=colors, quiet=True) # rename to requested output name g.rename(raster=(input_name, output_name), quiet=True) return output_name