-Bold-faced names refer to routines within PC ARC/INFO, ERDAS, or DOS.

-Italicized names refer to files (included here to facilitate following file lineage).

-Small print shows program options used in the described step.

Rasterize PC ARC/INFO wetland files with PC ARC/INFO POLYGRID.

Copy imagery in ERDAS format to 8mm tape at Adirondack Park Agency.

Copy imagery from 8mm tape to SUNY-P with TDRAW and DOS Copy (binary files option).

Use PC ARC/INFO-ERDAS Live Link, CURSES, MAKEFIL, SUBSET, and FIXHED to create a rasterized wetland file subset (SWWTL), a May LANDSAT imagery subset and a July LANDSAT imagery subset that possess the identical size (both pixel size and file size), file and real world coordinates.

Combine (with ERDAS MAKEFIL and SUBSET) May and July LANDSAT subsets into one 12-band subset with only the May and July reflective bands (i.e., bands 1, 2, 3, 4, 5, and 7).

MASK out wetlands from the 12-band image file with SWWTL to create SOBXWTL.

Use DIGSCRN to take signature samples from SOBXWTL and check within and between class signatures using ELLIPSE and CMATRIX. Extract signature statistics with SIGEXT. Take additional signature samples as necessary. Regroup signatures with SIGMAN and use for image classifications.

MAXCLAS SOBXWTL July reflective bands to JULYVEG with the classes: Deciduous, Deciduous/Open, Mixed, Conifer, and Open. Maximum likelihood classifier, classify 0's (wetland was also included as a class), no first-pass parallelepiped classification, no a priori values assigned.

THRESH JULYVEG(interactive selection). Compare THRESH percentages by showing histograms over Old Forge, Brantingham, and Eagle Bay; accept all classes. MASK SOBXWTL with Conifer and Open from JULYVEG and classify this subset to JULXVEG with classes Conifer, Cloudshadow07, Open, Barren, UrbanOpen, UrbanMix, and Cloud07. Maximum likelihood classifier, do not classify 0's, no first-pass parallelepiped classification, no a priori values assigned. Evaluate classes using CLASOVR and RECODEd Open and Barren to OpenVegetated; UrbanOpen and UrbanMix to Open (not vegetated). THRESH JULXVEGand choose class boundaries by showing histograms over Old Forge, Brantingham, and Eagle Bay. THRESH Conifer 1.15% to Cloudshadow07 (JULXVEGA) and Cloudshadow07 50.4% to Conifer (JULXVEGB).

OVERLAY JULYVEG + JULYXVEG + JULXVEGA + JULXVEGB + SWWTL to JULVEGA. RECODE to match accuracy assessment file pixels. Use POLYCAT and CLASSERR to assess accuracy.

RECODE JULYUPLA Wetland class to equal 0 (background). SCAN JULYUPLA to JULYUPLB to remove "salt-and-pepper" (3X3 rectangular window with the majority scan option, initialization border option with value = 0, majority threshold = 1, exclude 0 input values from analysis). SEARCH JULYUPLB to JULCLDA for Cloud07 and CloudShadow07 with a 3-cell search. RECODE JULCLDA to JULCLDB (0, 1, 2, 3 = 1 (July cloud/cloudshadow); 4 = 2 (no July cloud/cloudshadow)). MASK the 12-band image file (SWOBXWTL) with JULCLDB to JULCLDIM; only areas obscured by July clouds/cloud shadow remain.

MAXCLAS JULCLDIM May reflective bands to MAYUPA with classes Deciduous, Deciduous/Open, Mixed, Conifer, and Open. Maximum likelihood classifier, do not classify 0's, no first-pass parallelepiped classification, no a priori values assigned. Determine THRESH values for MAYUPA by interactive histogram evaluations over SWOBXWTL in Old Forge, Brantingham, and Eagle Bay. THRESH Deciduous/Open by 20.05% to Cloud05 (CLOUD05) and Conifer by 12.22% to Cloudshadow05 (CLDSHD05).

MASK JULCLDIM with MAYUPA Open to MAYUPOPA. MAXCLAS MAYUPOPA May reflective bands to MAYUPOPB with classes: Open, Barren, UrbanOpen, UrbanMix, and Cloud05. Evaluate MAYUPOPB with THRESH and CLASOVR in Brantingham (SWOBXWTL); too few pixels for evaluation in Old Forge and Eagle Bay. Cloud05 includes a bad spot in the imagery east of Brantingham. RECODE MAYUPOPB to MAYUPOPC and reclass Open, Barren, and UrbanMix to OpenVegetated; UrbanOpen to Open (not vegetated); and Cloud05 to Cloud (i.e., July and May cloud).

OVERLAY JULYUPLB + CLOUD05 + CLDSHD05 + MAYUPOPC with a majority option and RECODE to combine classes to create JULYUPLC. SCAN JULYUPLC to JULYUPLD(3X3 rectangle, majority option, border option: initialization with 0, exclude 0 values from input).OVERLAY JULYUPLD + SWWTL to form JULYUPLE. RECODE JULYUPLE to perform accuracy assessment with POLYCAT and CLASERR.

OVERLAY JULYUPLE + SWWTL to form SOBUPLND which isolates the Oswegatchie-Black study area. MASK SWOBXWTL + SWWTL, forming SOBXWTL - a 12-band image file of the study area with wetlands masked.

Rasterize PC ARC/INFO wetland files with PC ARC/INFO POLYGRID.

Copy imagery in ERDAS format to 8mm tape at Adirondack Park Agency.

Copy imagery from 8mm tape to SUNY-P with TDRAW and DOS Copy (binary files option).

Use PC ARC/INFO-ERDAS Live Link, CURSES, MAKEFIL, SUBSET, and FIXHED to create a rasterized wetland file subset (NWWTL), a May LANDSAT imagery subset and a July LANDSAT imagery subset that possess the identical size (both pixel size and file size), file and real world coordinates.

Combine (with ERDAS MAKEFIL and SUBSET) May and July LANDSAT subsets into one 12-band subset with only the May and July reflective bands (i.e., bands 1, 2, 3, 4, 5, and 7).

MASK wetlands from the 12-band image file with NWWTL to create NWOBXWTL.

Use DIGSCRN to take signature samples from NWOBXWTL and check within and between class signatures with ELLIPSE and CMATRIX. Extract signature statistics with SIGEXT. Take additional signature samples as necessary. Regroup signatures with SIGMAN and use for image classifications.

MAXCLAS NWOBXWTL to UPL07A using just the July reflective bands and classes: Deciduous, Conifer, Mixed, OpenVegetated, Barren, Cloud07, and Cloudshadow07. Maximum Likelihood classifier with no classification of 0's, no first pass parallelepiped classification, and the creation of a probability file. Check classification with THRESH overlaying the NWOBXWTL image of Newton Falls and Oswegatchie. All classes accepted.

SCAN UPL07A to UPL07B with a 3X3 rectangle to minimize "salt-and-pepper" pixels (majority option, border option = initialization with value = 0, majority threshold = 1, exclude 0 values from analysis). SEARCH UPL07B Cloud07 and Cloudshadow07 plus 3 pixels to UPL07C then RECODE to JULCLDNA (0, 1, 2, 3 = 1 (July cloud/cloud shadow); 4 = 2 (no July clouds/cloud shadow)). MASK the 12-band image file NWOBXWTL with JULCLDNA to create JCLDMSKN, preserving only the imagery beneath the July cloud/cloudshadow.

MAXCLAS JCLDMSKN to JULCLDNB with the classes Deciduous, Conifer, Mixed, OpenVegetated, Barren, Cloud05, and Cloudshadow05. Maximum Likelihood classifier, no processing of 0's, no first pass parallelepiped classification, no a priori values, and request the creation of a probability file. Visually check classification over NWOBXWTL May image using CLASOVR and discover that some Cloudshadow05 should have been classed as Conifer. THRESH 91.16% of Cloudshadow05 to Conifer (MAYCONIF).

OVERLAY UPL07A + JULCLDNB + MAYCONIF to UPL07C with maximum value dominating. RECODE UPL07C to UPL07D to regroup classes and prepare for accuracy assessment. SCAN UPL07D to UPL07E (majority option, 3 X 3 rectangle, do not scan 0s, initialize border with 0) to remove any "salt-and-pepper" pixels introduced with the May classification of July cloud cover. OVERLAY UPL07E + NWWTL to cut out study area only and RECODE it to match SOBUPLND, creating NOBUPLND. MASK NOBXWTL image file with NWWTL to isolate the study area, creating NOBXWTL.

Select accuracy assessment pixels using DIGSCRN on the Number Four, Newton Falls, Harrisville, Beaver River, Eagle Bay, and Big Moose quadrangles. Use POLYCAT and CLASERR to perform accuracy assessment.

READ SOBXWTL and NOBXWTL and visually check overlapping areas of SOBUPLND and NOBUPLND with CLASOVR.

STITCH SOBUPLND + NOBUPLND to form UPLANDA. MAKEFIL and SUBSET NOBUPLND to create NOBUPLA, the northern land cover GIS file with file dimensions of UPLANDA. OVERLAY NOBUPLA onto UPLANDA to UPLANDB with the northern file Deciduous and Conifer dominating all southern classes except Deciduous/Open and Wetland from both files superseding all other classes. RECODE UPLANDB to group classes, creating OBUPLAND.

Classification accuracy files for northern and southern land cover files updated with DOS EDIT to correspond with OBUPLAND class values and appended together with DOS COPY (CLASSACC). DOS COPY and EDIT used to separate CLASSACC into two separate files (CLASSACA, CLASSACB) to accommodate ERDAS limitations for a 512 maximum number of classification accuracy pixels per file. Run POLYCAT and CLASERR on both CLASSACA and CLASSACB and summarize results manually.