Handle conversion for very large rasters to prevent Overflow Error!

polygonToRaster.py

import itertools
import rasterio as rs
import geopandas as gpd
from rasterio.crs import CRS
from rasterio import features

def makeAffine(tf, lbl, w, h):
    if lbl == 'A':
        return tf
    if lbl == 'B':
        return tf * tf.translation((w/2),0)
    if lbl == 'C':
        return tf * tf.translation(0,(h/2))
    if lbl == 'D':
        return tf * tf.translation((w/2),(h/2))

lakes = gpd.read_file('./ohSevenA.shp')
src = rs.open('./fdr')
if src.crs != lakes.crs:
    lakes.to_crs({'init': u'epsg:5070'}, inplace=True)
meta = src.meta.copy()
w, h = src.width, src.height
meta.update(compress='lzw',
            nodata=0,
            dtype=rs.uint32,
            driver='GTiff',
            height = h/2,
            width = w/2,
            crs=CRS({'init': u'epsg:5070'}))

col_chunks = [(0, w/2), (w/2, w)]
row_chunks = [(0, h/2), (h/2, h)]
for win, lbl in zip(itertools.product(row_chunks, col_chunks),['A','B','C','D']):
    nuAff = makeAffine(src.transform, lbl, w, h)
    meta.update(transform=nuAff)
    lksRas = rs.open("./lakes_%s.tif" % lbl, 'w', **meta)
    arr = lksRas.read(1)
    shapes = ((g,v) for g,v in zip(lakes.geometry,lakes.COMID))
    burned = features.rasterize(shapes=shapes, fill=0,
                                out=arr,
                                out_shape=arr.shape,
                                transform=nuAff)
    lksRas.write(burned,indexes=1)
lksRas = None