APPENDICES

Appendix A:
Standard Reference Material Certification

Appendix B:
Data

Appendix C:
Data Analysis

Appendix D:
Sections and Plans
(Some Maps Exceed 11 x 17, not included in Electronic Filing)

SUMMARY

NovaGold Resources Inc. (NovaGold) has asked MRDI, a division of AMEC E&C Services Limited (MRDI), to provide an independent Qualified Person’s Review and Technical Report for the Donlin Creek project in Alaska. Stephen Juras, P.Geo., an employee of MRDI, served as the Qualified Person responsible for the preparation of the Technical Report as defined in National Instrument 43-101, Standards of Disclosure for Mineral Projects, and in compliance with Form 43-101F1 (the “Technical Report”).

The Donlin Creek project consists of gold mineralization in numerous deposits that are part of an early Tertiary age As-Sb-Au-Hg hydrothermal system located in southwest Alaska, approximately 70 km northeast of Aniak. The gold-bearing deposits have been combined into two resource areas: ACMA (containing ACMA proper and 400 Zone deposits) and Lewis (comprising Lewis or North Lewis, South Lewis, Vortex, Rochelieu and Queen deposits). The project is serviced by commercial air services out of Anchorage and Aniak and by a 25 km long winter road from the town of Crooked Creek.

Information and data for the review and report were obtained from NovaGold personal and the project site in Alaska. Stephen Juras conducted the site visit from October 26 to November 1, 2001.

The property consists of 109 km² (42 mi², 10,900 ha) of privately owned Native land. Calista Corporation, a regional Native corporation (Calista), owns the subsurface rights, and The Kuskokwim Corporation, a village corporation (TKC), owns the surface rights. Placer Dome (Placer) acquired a 20-year lease from Calista effective May 1, 1995. Annual property payments are U.S.$200,000 through the end of the lease. Annual work commitments are U.S.$1,000,000, but excess expenditures are carried forward. Calista holds a retained net royalty of 1.5% until payback of capital, increasing to 4.5% thereafter. NovaGold entered into an agreement with Placer in early 2001 to earn a 70% interest in the project by expending U.S.$10 million on exploration and development within ten years.

The Donlin Creek project geology consists of flysch sequence sedimentary units of the Cretaceous Kuskowim Group that have been intruded by Late Cretaceous to early Tertiary felsic intrusive rocks. The sediments consist of interbedded greywacke, shale and siltstone. Greywacke is dominant (Lewis resource area), but shale-rich areas also occur (ACMA resource area). The overall bedding strikes NW and dips 10° to 50° SW. The intrusive units consist of porphyritic rhyodacite and rhyolite and lesser mafic dykes and sills. Sills and dykes are common in the ACMA and southern Lewis areas, whereas dykes dominate in the areas to the northeast. The dykes and sills range from a few meters to more than 60 m in width.

Mineralization is best developed in the felsic intrusive rocks, with lesser mineralization in sediments (and then mainly greywacke). It is structurally and lithologically controlled along NNE-trending fault zones and best developed where those zones intersect favourable host lithologies such as the felsic intrusive dykes and sills and greywacke. The mineralization occurs as both disseminated zones (intrusive units only) and quartz, carbonate and sulphide (pyrite, arsenopyrite and stibnite) vein and veinlet networks. Native arsenic and realgar are also observed. The gold occurs primarily in the lattice structure of arsenopyrite.

The database used to estimate the mineral resources for the ACMA and Lewis deposits consists of samples and geological information from 133 and 331 drill holes, respectively. NovaGold drilled 42 holes in the ACMA resource area in 2001. The rest of the holes were drilled by Placer from 1995 to 2000. The drill hole database includes 14 reverse circulation (RC) holes at ACMA and 92 RC holes at Lewis. Data from these holes were used in the respective mineral resource estimates. Samples from 2001 were prepared at site and sent for analyses to Bondar-Clegg laboratory in North Vancouver, B.C. Data transfer to the resource database was validated from original assay certificates through a 5% check of the database.

Two sets of similar protocols were used for the samples that formed the basis of the Lewis and ACMA mineral resource models. Most of the samples from Placer’s work were processed in Placer’s own laboratory. NovaGold’s samples were processed by Bondar-Glegg, a commercial laboratory. However, the results can be evaluated together because the Standard Reference Material (SRM), the blank material and the duplicate protocol were the same. The performance of both SRMs was within acceptable limits and showed that the overall assay process was in control for the work done. However, performance of both standards did suffer somewhat in the most recent (2001) batches. Coincident results of both SRMs in some of the batches from that program fell below the 95% confidence limit. These deviations could indicate a lower-grade bias in samples represented by those batches. As a result, the ACMA resource model could be slightly conservative for material around the cut-off value. The duplicate sample program showed no bias in the assay process; performance was 90% within ±20% for values greater than 0.15 g Au/t. Good reproducibility of the gold values is demonstrated. The blank sample program worked well and demonstrated negligible contamination in the assay process.

The mineral resource estimates for the Donlin Creek project were calculated under the direction of Stephen Juras. The estimates were made from 3-dimensional block models utilizing commercial mine planning software (MineSight®). Industry-accepted methods were used to create interpolation domains based on mineralized geology and grade estimation based on ordinary kriging. For both areas, acceptable mineralized envelops were defined through Probability Assisted Constrained Kriging, or PACK, on gold thresholds of 0.7 g Au/t and 0.5 g Au/t for ACMA and Lewis, respectively. This method best limited the waste intervals of the intrusive units at ACMA from diluting the

grades in the mineralized regions and best honoured the significant contribution of greywacke-hosted mineralization together with mineralized felsic intrusive units at Lewis. Extreme high gold grades were capped (in the 2 m composite database) to 30 g Au/t and 20 g Au/t for ACMA and Lewis, respectively.

Reasonableness of grade interpolation was reviewed by visual inspection of sections and plans displaying block model grades, drill hole composites and geology. Good agreement was observed. Global and local bias checks in block models, using nearest-neighbour estimated values versus the ordinary kriged values, found no evidence of bias.

The logic for mineral resource classification of ACMA and Lewis was consistent with the CIM definitions referred to in National Instrument 43-101. The Indicated mineral resource category is supported by the present drilling grids over the ACMA and Lewis deposits (nominal 25 m to 35 m). The Measured mineral resource category is supported only in localized areas of the two deposits that have a nominal drill grid spacing of about 15 m.

The mineralization of the Donlin Creek project as of January 24, 2002, is classified as Measured, Indicated and Inferred Mineral Resources. The classified Mineral Resources are shown in Table 1.1. MRDI and NovaGold selected cut-off grades of 3.5 g Au/t, 2.5 g Au/t and 1.5 g Au/t as being representative of the large-scale open pit mining operation that would potentially be economic for gold prices of U.S.$250, U.S.$300 and U.S.$350 per ounce of gold, respectively.

The 2002 mineral resource estimates for Donlin Creek project show a marked increase over historical estimates. This increase is the cumulative result of: additional drilling at ACMA, re-interpretation of the felsic intrusive units (particularly the sills), continued evolution of understanding the structural controls on the gold mineralization, and the use of PACK envelops (see above) to best honour the contribution of greywacke-hosted mineralization together with mineralized felsic intrusive units at Lewis.

The 2002 mineral resource estimates for Donlin Creek project show a marked increase over historical estimates. This increase is the cumulative result of: additional drilling at ACMA, re-interpretation of the felsic intrusive units (particularly the sills), continued evolution of understanding the structural controls on the gold mineralization, and the use of PACK envelops (see above) to best honour the contribution of greywacke-hosted mineralization together with mineralized felsic intrusive units at Lewis.

INTRODUCTION AND TERMS OF REFERENCE

NovaGold Resource Inc. (NovaGold) has asked MRDI, a division of AMEC E&C Services Limited (MRDI), to provide an independent Qualified Person’s Review and Technical Report for the Donlin Creek project in Alaska. Stephen Juras, P.Geo., an employee of MRDI, served as the Qualified Person responsible for the preparation of the Technical Report as defined in National Instrument 43-101, Standards of Disclosure for Mineral Projects, and in compliance with Form 43-101F1 (the “Technical Report”).

Information and data for the review and report were obtained from NovaGold personal and the project site in Alaska.

The work entailed review of pertinent geological data in sufficient detail to prepare the Technical Report. Stephen Juras, P.Geo., in addition to supervising the preparation of the Technical Report, conducted and supervised the review of the geological data and the resource estimation work by NovaGold and MRDI. Stephen Juras also conducted the independent site visit to the Donlin Creek project.

Terms of Reference

This report addresses two main areas of mineralization, each having defined sub-zones, as follows:

DISCLAIMER

No disclaimer statement was necessary for the preparation of this report.

PROPERTY DESCRIPTION AND LOCATION

Donlin Creek is located in southwest Alaska, approximately 70 km (44 mi) northeast of Aniak, a regional hub (see Figure 4.1). The property consists of 109 km² (42 mi², 10,900 ha) of privately owned Native land. Calista Corporation, a regional Native corporation (Calista), owns the subsurface rights, and The Kuskokwim Corporation, a village corporation (TKC), owns the surface rights. The resource areas are within T. 23 N., R. 49. W. (see Figure 4.2), Seward Meridian, Kuskokwim Recording District, Crook Creek Mining District, Iditarod A-5 USGS 1:63,360 topography map. These areas consist of the ACMA and 400 Zone prospects (grouped as ACMA) and the Lewis, South Lewis, Vortex, Rochelieu and Queen prospects (grouped as Lewis) (see Figure 4.3).

Mineral Tenure

The land status of the Donlin Creek area is shown in Figure 4.2. Most of the rights (surface and subsurface) are governed by conditions defined by the Alaska Native Claims Settlement Act (ANCSA). Section 12(a) of ANCSA entitled each village corporation to select surface estate land from an area proximal to the village in an amount established by its population size. Calista receives conveyance of the subsurface when the surface estate in those lands is conveyed to the village corporation. Section 12(b) of ANCSA allocated a smaller entitlement to the regional corporations with the requirement they reallocate it to their villages as they choose. Calista receives subsurface estate when its villages receive 12(b) lands. Calista reallocated its 12(b) entitlement in 1999 according to a formula based on original village corporation enrolments.

TKC and Calista received patent to the 12(a) surface and subsurface estate, respectively, at Donlin Creek and Crooked Creek, shown in yellow in Figure 4.2. They also have 12(a) and 12(b) selections in the areas shown in yellow with a dashed blue border; these have been prioritized for conveyance pending completion of surveys and easement identification. TKC and Calista have committed to taking conveyance of all their selections in the Crooked Creek withdrawal area, which includes Donlin Creek.

ANCSA Subsection 14(h)(8) allows regional corporations to select federal lands for conveyance to fee simple land, both surface and subsurface estate. Lands selected and conveyed under this section of the Act are called 14(h)(8) lands.

The in-lieu lands are selections at this time. These are subsurface selections Calista made in lieu of the subsurface estate it was prohibited from owning when villages were conveyed land within the wildlife refuges that predated ANCSA.

The allotments are private land transferred directly to individual Alaska Native applicants from the federal government, based on use and occupancy. The Native Allotment Act of 1906 pre-dated and was extinguished by ANCSA, and allotees have priority over ANCSA selections. There are no Native allotments on Crooked Creek beyond the local village area or in the Donlin Creek area.

None of the state land has been Tentatively Approved or patented to the state. The state made all its selections in this area after 1991, and the land remains under the ownership and administration of the BLM until future conveyance.

Permits and Agreements

Placer Dome (Placer) acquired a 20-year lease from Calista effective May 1, 1995. Annual property payments are U.S.$200,000 through the end of the lease. Annual work commitments are U.S.$1,000,000, but excess expenditures are carried forward. Calista holds a retained net royalty of 1.5% until payback of capital, increasing to 4.5% thereafter. NovaGold has entered into an agreement with Placer to earn a 70% interest in the project by expending U.S.$10 million on exploration and development within ten years. Once the financial commitment has been fulfilled, Placer has 90 days to decide on one of three options: a) to remain at 30% interest and participate as a minority partner; b) to convert to a 5% Net Profits Interest (NPI); or c) to exercise a back-in right to re-acquire a majority interest in the project (70%) by expending three times the amount expended by NovaGold at the time the back-in is exercised; conducting a feasibility study; and making a decision to mine at a production rate of not less than 600,000 ounces of gold per year within a five-year period from the exercise back-in. For the third option, NovaGold would contribute its share of costs after Placer has expended three times NovaGold’s initial earn-in expenditure. Calista also has a back-in provision in the project where it may acquire 5% to 15% interest in the deposit by providing its share of accrued capital costs.

Placer has implemented all of the necessary permits for exploration and camp facilities. These permits are active at the Alaska Department of Natural Resources (hard rock exploration, temporary water use), the Corp of Engineers (individual 404 and nationwide 26,), Alaska State Department of Conservation (wastewater, drinking water, food handling), the Alaska Department of Fish and Game (title 16 – fish), the Environmental Protection Agency (NPDES) and the Federal Aviation Administration (airport). NovaGold has maintained and abided by the stipulations of these permits.

ACCESSIBILITY, CLIMATE, LOCAL RESOURCES, INFRASTRUCTURE AND PHYSIOGRAPHY

The Donlin Creek property is located in southwestern Alaska, approximately 19 km (12 mi) north of the village of Crooked Creek on the Kuskokwim River (see Figure 4.1). The Kuskokwim River is a regional transportation route and is serviced by commercial barge lines. A 25 km (15 mi) long winter road, designated as an Alaska State Highway route and transportation corridor, accesses the property from the barge site at the town of Crooked Creek (Figure 4.2). The project has an all-season camp with facilities to house up to 75 people and an adjacent 1,500 m (5,000 ft) long airstrip capable of handling aircraft as large as C-130 Hercules (19,050 kg or 42,000 lb capacity), allowing efficient shipment of personnel, large equipment and supplies. The project is directly serviced by commercial air services out of both Anchorage, 450 km (280 mi) to the east, and Aniak, 80 km (50 mi) to the west.

The project area is one of low topographic relief on the western flank of the Kuskokwim Mountains. Elevations range from 150 m to 640 m (500 ft to 2,100 ft). Ridges are well rounded and easily accessible by all-terrain vehicle. Hillsides are forested with black spruce, tamarack, alder, birch and larch. Soft muskeg and discontinuous permafrost are common in poorly drained areas at lower elevations.

The area has a relatively dry interior continental climate with typically less than 50 mm (20") total annual precipitation. Summer temperatures are relatively warm and may reach nearly 30°C (83°F). Minimum temperatures may fall to well below -20°C (0°F) during the cold winter months.

The project is currently isolated from power and other public infrastructure and will operate as a stand-alone project. Sufficient space is available to site the various facilities, including personnel housing, stockpiles and processing plants. Ample water supply is available from surface and subsurface sources.

HISTORY

Previous operators have undertaken significant work on the property. Table 6.1 summarizes the work history at Donlin Creek. It is interesting to note that no placer gold mining has been done in American Creek, the stream draining the Lewis and ACMA area. The micron-sized gold found in these lode deposits is probably not conducive to placer concentration. The lode deposits in the Dome area (Figure 4.3) contain free coarse gold, and economic placer deposits have been developed downstream.

1996 Work

Major activities included:

1998 Work

The main tasks completed in 1998 include:

Regional Geology

The Cretaceous Kuskokwim Group, a post accretionary basin-fill flysch sequence, is the dominant unit found in the region (Figure 7.1). Estimated to be up to 12 km thick, the group consists primarily of lithic sandstone (greywacke) with lesser interbeds of carbonaceous shale and siltstone. Generally the Kuskowim Group displays no penetrative metamorphic fabric but does contain localized zones of open to isoclinal folds.

The Kuskokwim Group has been intruded and locally overlain by Late Cretaceous to early Tertiary intrusions, dykes, sills and subaerial volcanic rocks. Many of the dykes appear to have been emplaced along or near NE-trending fault zones. Plutonic rocks comprise monzonite, quartz monzonite, syenite, granodiorite and granite, and both intrude and are overlain by coeval volcanic rocks. Porphyritic rhyodacite and rhyolite dykes and sills occur throughout the region. Contacts between the igneous dykes and sedimentary rocks of the Kuskowim Group are typically sharp and without hornfelsic margins. K-Ar ages indicate two intrusive events, one around 71 Ma and the other around 61 Ma.

The Donlin Creek project area lies between two major NE-trending right lateral faults found in southwest Alaska: the Denali-Farewell fault system to the south and the Iditarod-Nixon Fork fault system to the north (see Figure 7.1). The region contains abundant NE to ENE- and NW to WNW -trending lineaments that likely represent steeply dipping strike slip faults. Displacement along the main faults in the Donlin Creek regions is inferred to be right-lateral on NE structures and left lateral on NW faults. Because of the paucity of outcrop along the main faults in the region, the inferred location and sense of displacement are speculative.

Property Geology

The Donlin Creek property geology is illustrated in Figure 7.2. The main rock types are greywacke, shale and siltstone of the Kuskowim Group. Greywacke is dominant (Lewis deposits) but shale-rich areas are common (ACMA deposits). The overall bedding strikes NW and dips 10° to 50° SW. Numerous dykes and sills intrude the Donlin Creek sedimentary rocks, with the bulk of the igneous units occurring in a NE-trending corridor about 8 km long. The dykes and sills range from a few metres to more than 60 m in width and are composed of porphyritic rhyodacite and rhyolite and lesser mafic units. Sills and dykes are common in the ACMA and southern Lewis areas, whereas dykes dominate in the areas to the northeast.

The porphyritic rhyodacite has been divided into five field units (RDX, RDA, RDF, RDL, RDB) of similar composition and mineralogy. These units are temporally and spatially related and reflect textural and perhaps temporal variations of a related intrusive phase. Differences include amount and size of phenocrysts, fineness of groundmass and overall colour. The most abundant and thickest phase is the RDX, which is characterized by abundant large phenocrysts (about 30%) in a fine-grained crystalline groundmass. RDA refers to intrusive phases that contain somewhat smaller and less-abundant phenocrysts in an aphanitic groundmass. RDL is a unique-looking RDX-like unit, but the phenocrysts are commonly as coarse lath-like crystal. RDF is a fine-grained porphyritic felsic unit with fine phenocrysts in an aphanitic goundmass. In parts of the ACMA, it is interpreted to represent an early phase of the intrusive activity. RDB is sometimes called blue porphyry because of its dark grey to bluish grey groundmass.

Mafic dykes occur throughout the property. They are less common than the feslic intrusive rocks and generally are highly altered. They appear to be the oldest intrusive phase in the property.

High- and low-angle faults trending NNE and NW reflect the dominant structural trends. These major structures are clearly evident in the property geology and aeromagnetic data. Mineralization is structurally and lithologically controlled along NNE-trending fault zones and is best developed where those zones intersect favourable host lithologies such as felsic intrusive dikes and sills and greywacke. Within the intrusive sills, the hinges of an east-west trending antiformal structure appear to be spatially related to the highest-grade mineralization, especially where these structures are cut by the NNE-trending faults. The mineralized trend is marked by a pronounced NE-trending aeromagnetic low that is related to a low magnetic signature in the intrusive rocks, magnetite destructive alteration and thermal metamorphism of the surrounding sedimentary rocks.

Deposit Geology

The ACMA/400 deposits consist of multiple felsic sill phases that intruded a dominantly shale sedimentary sequence (Figure 7.3). All five intrusive phases are present. A thrust fault cuts the deposits and appears to have played a role in controlling intrusive extent. Immediately below this fault, the sills thicken significantly. No corresponding thickening occurs in the hanging wall to this fault. The implication is that the intrusive event and thrust fault formation were coeval. The thrust fault acted as a barrier to the ongoing intrusive event, pooling the intrusive material against it. The mineralization is concentrated within the intrusive units in this region.

The Lewis group of deposits consists of multiple felsic dyke phases (northern area, Figure 7.4) to sill and dyke phases (southern area, Figure 7.5) that intrude a greywacke, shale sequence. The appearance of sills could denote the presence of another thrust-like structure similar to that identified at ACMA. The mineralization is found almost equally in the intrusive units and the greywacke.

DEPOSIT TYPES

Two separate gold-rich systems are recognized along the Donlin Creek trend. One is an early, high-temperature porphyry style at the Dome prospect that contain free Au with trace Ag, Cu, Zn, Bi, Te and Se. A later and lower-temperature, low-sulfidation epithermal system was responsible for most of the identified mineralization on the Donlin Creek trend. It is seen in the Lewis, ACMA and Snow areas. This main mineralizing system contains Au within arsenopyrite and has an Au-As-Sb-Hg geochemical signature. Mineralization is best developed in intrusive rocks, lesser so in sediments (and then mainly greywacke). It is present as both disseminated zones and quartz, carbonate and sulphide (pyrite, arsenopyrite and stibnite) vein and veinlet networks. Native arsenic and realgar are also commonly observed. The bulk of the gold occurs in the lattice structure of arsenopyrite.

Alteration from the main Donlin Creek mineralizing system includes large zones of illite-quartz-pyrite alteration found largely within intrusive rocks surrounding mineralized zones. Typically both matrix and feldspar phenocrysts are strongly altered to illite, but kaolinite and illite+kaolinite also occur. This grades outward into relatively weakly altered intrusive with interlayered chlorite/smectite and minor kaolinite, and illite and carbonate alteration. Fluid inclusion studies indicate filling temperatures for these samples at 150°C to 220°C.

MINERALIZATION

Gold mineralization at Donlin Creek is lithologically and structurally controlled. Disseminated mineralization occurs in rhyodacite dikes. Structurally controlled mineralization is along NNE-trending extensional fracture zones and is developed best where these zones intersect competent lithologies such as felsic dykes and sills, or massive greywacke. Quartz -carbonate-sulphide (pyrite, stibnite and arsenopyrite) veins are the primary mineralized features, but gold mineralization also occurs in thin, discontinuous veinlets and fracture fillings. Veinlets seldom exceed 1 cm in diameter, and most fracture fills are just thin sulphide coats on fracture surfaces. Gold is sometimes associated with realgar/orpiment as late-stage fracture filling in sediments and intrusive rocks and disseminated in intrusive rocks.

Ore and related minerals include pyrite, marcasite, arsenopyrite, stibnite, realgar, orpiment and limonite, with lesser amounts of native arsenic, native gold, cinnabar, covellite, chalcopyrite, galena, pyrrhotite, malachite, sphalerite, scorodite, stibiconite?, kermesite? and hematite. Trace amounts of enargite were seen in thin section enclosed in an arsenopyrite grain, and trace amounts of fine-grained gersdorffite were observed in mafic dykes.

Pyrite is common and appears to be the earliest sulphide phase. It is ubiquitous in the rhyodacite phases and occurs as disseminated grains and microfracture fillings. Disseminated pyrite in the sedimentary rocks occurs as fine to coarse grains (up to 5 mm across) preferentially concentrated near dyke/sill contacts. Relative abundance of pyrite is not an indicator of gold grade.

Arsenopyrite is deposited later than pyrite, commonly replacing the pyrite. Arsenopyrite occurs as fine to very fine grains that are disseminated in intrusive rocks and occur as fracture/vein fillings. Fine-grained arsenopyrite is generally difficult to distinguish from ubiquitous disseminated graphite, causing visual estimates made on drill core to vary widely.

Native arsenic occurs as dark grey, granular massive to reniform grains commonly associated with stibnite in dolomite veinlets. Stibnite commonly occurs as disseminated grains and masses within carbonate veins, and occasionally as interlocking needles up to 2 mm x 2 cm to 3 cm in open spaces within quartz-carbonate veins and on fracture surfaces.

Gold seen in polished thin sections occurs as 1 µm to 3 µm blebs with no clear paragenetic relationship to other minerals. Metallurgical tests indicate that 95% to 98% of the gold in the Lewis area is contained in arsenopyrite, and that fine-grained arsenopyrite (<20 µm diameter) contains 5 to 10 times more gold than coarse-grained arsenopyrite.

Alteration patterns associated with the Donlin Creek mineralization are not complex. All intrusive units are altered and fresh rocks are uncommon. Sericite dominant alteration (sericite+illite+kaolinite+carbonates+pyrite) is pervasive but varies in intensity. Most sericite is a whitish colour, but some plagioclase phenocrysts are replaced by light-green sericite, especially in weakly altered intrusive rocks. Illite and sericite have been identified by XRD. Pervasive carbonate alteration (mostly dolomitic composition) appears to be a separate alteration event that overprints sericite alteration. Silicification is not a significant event at Donlin Creek. Argillic/clay alteration is not present, but minor clay can be associated with the sericite alteration assemblage and within fault zones. Carbonate alteration is found in the mafic dykes where it can replace almost the entire rock.

Very fine grained graphite frequently occurs in all igneous units. The graphite is thought to be an alteration product, but it is not clear whether it is a separate alteration event or part of the sericite and carbonate alteration events. Graphite is common and occurs in open spaces or high porosity areas, often with coarse sericite or as fine, isolated shred-like fragments and grains.

Distinctions between grey and blue porphyry have been recognized as an alteration effect and not a primary igneous feature. The blue colour, when present, is stronger near faults and rhyodacite/sedimentary rock contacts. In several instances, the blue/grey boundary cuts feldspar phenocrysts and clearly relates to later alteration and/or mineralization events. The blue colour most likely is due to extremely fine, disseminated graphite in the rhyodacite matrix, or possibly finely disseminated sulphides.

EXPLORATION

NovaGold’s exploration work in 2001 consisted of diamond core drilling (7,403 m) and minor trenching (822 m) (see Figure 10.1). Drilling was concentrated in the ACMA/400 area and had two purposes: to in-fill known mineralized areas, focusing on higher-grade zones, and to extend known mineralized zones. The in-fill drilling was targeted to better understand the controls on mineralization and the geology (see Figures 7.3 to 7.5). This information was used to help locate the step-out drilling. Hole locations were based on a 25 m x 25 m drill grid and were inclined approximately at 50° in an attempt to intersect mineralization at as near perpendicular as possible.

NovaGold also re-examined existing geophysical data and soil geochemical data. These data were gathered during Placer’s five-year period of exploration in the project area. Soil geochemistry had proved to be a useful tool in this area. Figures 10.2 and 10.3 show a contoured representation of the data over the mineral resource area and the property as a whole. The anomalies served as a guide for past and current drill and trench programs, but more remain to be evaluated.

NovaGold geologists also re-mapped existing trenches over the Lewis and ACMA areas and re-logged previously drilled core.

2001 Drilling

NovaGold carried out diamond drilling in the ACMA/400 Zone area of the Donlin Creek project in 2001. Drilling totals 7,403 m in 42 drill holes. The holes generally range in length from 64 m to 404 m, averaging 176 m. The locations of the 2001 drill holes are shown in Figures 4.3 and 10.1.

Drilling was done by wireline method using H-size equipment (HQ). Two drill rigs were used, with one set up to retrieve oriented core for structural measurements. MRDI observed the operation of the oriented core drill rig during the site visit. Drilling was well supervised, the site was clean and safe, and work was efficiently done. Upon completion, the collar and anchor rods were trimmed to about 0.5 m above ground and capped.

Drill hole collars were located respective to a property grid. Proposed hole collars and completed collars were surveyed by triangulation using a Topcon GTS-211D total station and with a triple-prisms rod. Survey stations available were those established by Placer. Coordinates were given in the UTM coordinate system. MRDI checked three randomly chosen drill collars with a GPS unit. The results returned the recorded surveyed collar coordinates (within the ±4 m accuracy of the unit).

Most holes were drilled at a declination of between 50° and 70°; three holes were drilled vertically. Down-hole surveys were taken about every 45 m using a single-shot Sperry Sun instrument.

Standard logging and sampling conventions are used to capture information from the drill core. The core is logged in detail directly into a lap-top computer (data downloaded into the main Donlin Creek MS Access database daily). Four sets of data are captured in separate tables: Lithology, Mineral, Structural and Geotechnical. Any remarks are captured in a fifth table. Lithology is documented by a 2 to 4 letter alpha code (Table 11.1). The Mineral table captures visual percent veining (by type) and sulphide (pyrite, arsenopyrite, stibnite and realgar). Structural data consist of type of structure, measurements relative to core axis and oriented core measurement. The Geotechnical table records percent Recovery and RQD for the entire hole, and fracture intensity where warranted. The protocols and coding are similar to those used by Placer during its campaigns.

Pre-2001 Drilling

Most of the data for the Donlin Creek project is represented by drill holes drilled prior to 2001. The drilling was done by Placer from 1995 to 2000 (see Section 6). The ACMA resource area contains 91 pre-2001 drill holes, comprising 77 core holes (19,822 m, average length 257 m) and 14 RC holes (1,676 m, average length 120 m). The RC holes were drilled throughout the general area, with only 5 in the mineralized trend and none in the main ACMA deposit. The Lewis resource area is defined by 331 pre-2001 drill holes, comprising 239 core holes (60,054 m, average length 251 m) and 92 RC holes (10,113 m, average length 109 m). The RC hole coverage is concentrated over two deposits: Lewis proper or North Lewis (40 holes) and Queen (21). Also, no RC hole penetrated below 200 m elevation in any of the five deposits that make up the Lewis resource region.

MRDI examined the RC data and their use in the resource modelling process. The ACMA holes essentially were too few to have any substantial impact in the ACMA resource estimate. Those that were in the mineralized area were supported in terms of geology and gold grades by nearby core holes. As such, there was no apparent reason why the information should not be included.

The Lewis RC holes are more prevalent and will potentially have an impact on the Lewis resource estimate. Placer re-drilled many of the areas covered by RC holes with core holes. Visual inspection showed that in almost all cases, the RC and core geology and the relative intensity of gold mineralization supported each other. In places, the holes were close enough to examine for RC versus core bias. However, this test has to be implemented with caution because of the structural control of the mineralization (most of the mineralization follows steeply dipping structural zones). Ideally, this comparison should be made with the use of a structural model. Of the few sets of RC-core pairs that seemed usable, the results were mixed: some pairs showed no bias, others gave an apparent RC higher-grade bias and still others (slight majority) gave an apparent core higher-grade bias. Another cautionary note is that in re-drilling the RC grid, some of the holes may be examples of selection bias (where the apparent twinning of a high-grade zone will yield lower values and vice versa). As a consequence of this review, no obvious reason was seen to reject the RC data for use in the Lewis resource estimate.

MRDI observed during the geological and resource modelling process that drill hole sampling in the highest-grade ACMA and Lewis gold zones is by core holes only. Also, MRDI confirmed that core data solely outlined any deep gold mineralization that could potentially define ultimate open pit bottom areas.

SAMPLING METHOD AND APPROACH

The sampling protocol for the 2001 NovaGold program began with samples being outlined by the logging geologist. The entire hole was sampled, and the samples were usually marked at 2 m spacing. Sample intervals were broken at changes in lithologic units. This approach is sound and appropriate for this style of gold mineralization.

The core was then digitally photographed and sawn in half by diamond saw. Attention was paid to core orientation. One half was returned to the core box for storage at site and the other bagged for sample processing. The samples were processed at an on-site facility at the Donlin Creek campsite.

The procedure, witnessed by MRDI during its site visit, entailed the following steps:

The sample process lab is kept very clean and orderly, and all equipment appeared to be well maintained.

The samples were then sent to Bondar-Clegg’s laboratory in North Vancouver, B.C., for pulverization and gold analysis.

The sample protocol described above was one that Placer established and used to process the samples collected during its campaigns.

Significant composited assays for the Donlin Creek project are shown in Appendix B. Only values greater than 3 g Au/t are shown.

SAMPLE PREPARATION, ANALYSES AND SECURITY

Two sets of similar protocols were used for the samples that formed the basis of the Lewis and ACMA mineral resource models. Most of the samples from Placer’s work were processed in Placer’s own laboratory. NovaGold’s samples were processed by Bondar Clegg, a commercial laboratory. However, the results can be evaluated together because the Standard Reference Material (SRM), the blank material and the duplicate protocol were the same.

The samples for the NovaGold 2001 work were pulverized into a pulp (to better than 90% minus 150 mesh, or 100 µm) and analyzed by a 1-assay ton method, wherein a 29.17 g sub-sample was taken from the pulp sample, fire assayed and analyzed using an atomic absorption spectroscopy (AAS) finish. Samples that assayed 10 g Au/t or more were re-assayed by a gravimetric technique.

The protocol for samples from the Placer programs was very similar. A 25 g sub-sample was taken from the pulp, fire assayed and analyzed using an AAS finish. Samples that assayed 3 g Au/t or more were re-assayed by fire assay pre-concentration with a gravimetric finish.

Rigorous quality assurance programs were in place for virtually all the samples to be used in the ACMA and Lewis mineral resource estimate. Placer created four in-house control standard reference materials or standards. Two were used consistently throughout Placer and NovaGold’s work: Geological Gold Standard C and Geological Gold Standard D. These standards were made according to an accepted methodology of homogenization and round-robin assaying. The certification process was supervised by Placer’s assay team. Results of this work for these two standards are shown in Appendix A. One or both standards were inserted in all batches, depending on the by range of expected values. Graphs showing the performance of the standard analyses over the Placer and NovaGold work campaigns are provided in Figures 13.1 and 13.2.

The performance of both standards was within acceptable limits and demonstrated that overall the assay process was in control for the work done. The charts, however, indicate that performance of both standards did suffer somewhat in the most recent (2001) batches. Figure 13.3 shows the performance of the standards from the 2001 work. Coincident results below the lower 95% confidence limit for both standards occur for four to five batches in the middle of the program and over the last two batches of the completed work. These deviations outside the acceptance range of values for both standards could indicate a lower-grade bias in samples represented by those batches. During the quality control phase of a program, these results would

have triggered a request to the main laboratory to see if there was any readily apparent reason for the deviation. If no reason could be given, then a request to re-assay the suspect batches (including the control samples) should have been made. The risk associated with samples from these 2001 batches would likely be minimal. The potential low bias would affect areas around a cut-off grade threshold and if confirmed would mean that the estimate of material above a cut-off value in the ACMA model may be slightly conservative locally.

The duplicate samples taken for the Donlin Creek programs represent field or coarse reject duplicate samples and are used to evaluate the sample preparation and the analytical precision. Figure 13.4 shows the results in a relative difference versus relative mean grade chart. The scatter about 0% relative difference is symmetric, suggesting no bias in the assay process. Figures 13.5 to 13.7 display the same data in relative difference versus percentile rank charts. Results show that performance is 90% within ±20% for values greater than 0.15 g Au/t. As such, MRDI considers the duplicate program to have performed well and to indicate good reproducibility of the gold values.

The duplicate data were evaluated for relative precision using the Thompson-Howarth calculation method. Results are presented in Figure 13.8. The relative precision is

10% or better for gold grades above 0.7 g Au/t. The range of probable cut-off grades lies well into the flat (nearly horizontal) part of the precision versus grade curve.

Blanks were used to check for the presence of contamination in both sample preparation and assaying. Placer had collected a large container of crushed unmineralized diorite-like material for use as the blank material. Results are shown in Figure 13.9. Almost all values (99%) lie below 0.10 g Au/t and average of 0.013 g Au/t. MRDI examined several of the values that were higher than 0.10 g Au/t by checking their relevant assay certificates. Almost all values examined could be explained by sample switching (e.g., instruction sheet said to “put blank in front of sample x” but it was put behind and not noted in the instruction sheet). The obvious ones should be noted and corrected in the database. Nonetheless, the blank sample program worked well and demonstrates negligible contamination in the assay process.

DATA VERIFICATION

As a test of data integrity, the data used to estimate the January 2002 Donlin Creek mineral resources were checked several ways. MRDI initially conducted a 5% check of randomly chosen drill holes in each of the ACMA and Lewis regions and checked gold values against the original electronic assay certificates. No errors were uncovered. MRDI also checked the down-hole survey data. Camera shots were read for the check drill holes and compared to those stored in the resource database; a significant transcription error rate was found in all regions. As a result, NovaGold instituted a 100% check of the camera shot readings. MRDI re-checked the survey data after this work was completed and found no errors. Collar coordinates were checked against the database entries. Also, as mentioned in Section 11, MRDI checked three randomly chosen drill collars with a GPS unit. Readings obtained matched those entered in the database.

MRDI concludes that the assay and survey database used for the Donlin Creek mineral resource estimation is sufficiently free of error to be adequate for resource estimation.

ADJACENT PROPERTIES

Adjacent properties are not relevant for the review of the Donlin Creek project.

MINERAL PROCESSING AND METALLURGICAL TESTING

NovaGold is currently reviewing Placer’s metallurgical work (see Section 6) and initiating additional investigations on samples from selected drill core. Work is still in progress at the time of this report.

MINERAL RESOURCE AND MINERAL RESERVE ESTIMATES

The mineral resource estimates for the Donlin Creek project were calculated under the direction of Stephen Juras, P.Geo. The estimates were made from 3-dimensional block models utilizing commercial mine planning software (MineSight®). The project was divided into two resource areas: ACMA (ACMA and 400 zones) and Lewis (Lewis, South Lewis, Vortex, Rochelieu, Queen zones). The ACMA resource database comprised assay and geological data from 133 drill holes (119 core, 14 RC). The resource database for the Lewis model consisted of data from 331 drill holes (239 core, 92 RC). Projects limits are in truncated UTM coordinates. Project limits for ACMA are 39300 to 40900 East, 78000 to 79100 North, -400 m to +250 m elevation. Project limits for Lewis are 40900 to 42500 East, 78000 to 80000 North and 0 m to +450 m elevation. Cell size was 10 m east x 10 m north x 5 m high.

Geologic Models

Modelling philosophy was based on using geologic features and distribution of gold mineralization in controlling the extent of interpolation. The felsic dykes and sills are the main host for gold mineralization at Donlin Creek. However, only areas that have been structurally prepared within the intrusive rocks contain the gold-bearing mineralization. Additionally, the hosting sediments in ACMA and Lewis resource areas also can contain gold mineralization. In the shale-dominant ACMA area, the sediment-hosted mineralization is uncommon and is usually localized around the felsic intrusive contacts and high-angle structures. The Lewis area, by contrast, contains a significant greywacke component in the hosting sedimentary units. The gold distribution in the greywacke can be similar to the distribution in the felsic intrusive units. Commonly the mineralization passes from one unit to the other along a structurally prepared zone and, in places, is distributed in both without regard to the sediment-intrusive contact.

Interpolation is usually controlled by defining mineralized envelops. At Donlin Creek, the two resource areas posed two different challenges. For ACMA, NovaGold developed a revised felsic intrusive model. However, this model alone was not adequate as an envelop because the background value assays from large unmineralized portions of these intrusive units tended to unduly lower the grade of the mineralized areas in the interpolation process. At Lewis, a method was needed in which a mineralized zone was defined independently of lithology because of the significant contribution of greywacke-hosted mineralization. For both areas, acceptable mineralized envelops were defined through Probability Assisted Constrained Kriging, or PACK. PACK has been used by AngloGold at the Jerritt Canyon, Nevada, and Cripple Creek, Colorado, districts for many years. It has also been used at the Phoenix project and Hycroft mine, both in Nevada, and by Cambior at Cerro San Pedro, Mexico.

Interpolation approach was similar for each area. Drill hole data were composited into equal length composites of 2 m (honouring the geology as defined in Table 11.1). The composites were then separated into mineralized and non-mineralized groups by gold grade. Separation thresholds of 0.7 g Au/t and 0.5 g Au/t were chosen for ACMA and Lewis, respectively. Gold grade indicator variograms were calculated for each threshold and examined.

The probability that gold in any block in the model exceeded the threshold was estimated using the indicator data, the indicator variograms and ordinary kriging. For ACMA, the interpolation was constrained to intrusive units only, and a soft boundary was used at intrusive-sedimentary contacts on the search for composites so that any mineralized sediment values along the contacts could be incorporated. At Lewis, no lithologic constraint was applied. Contoured probability values (in increments of 0.05) defined the potential volumes used to delineate mineralized and non-mineralized populations. The boundary formed by each contour line was called an estimation envelope, or “shell.” The process of selecting an appropriate shell was iterated through several steps. The gold values from inside and outside any shell were extracted and examined. Plots of drill hole sections and the superimposed shell boundaries were also examined. The shell selected to outline the populations was based on these reviews. Finally, the review ensured the nominated boundaries did not violate current geologic understanding of mineralization controls. Final probability values used to define mineralized shells for grade estimation were 0.27 at ACMA and 0.28 at Lewis. Histograms and probability plots of composite gold data both inside and outside the PACK shells are shown in Appendix C.

Capping

Extreme high gold grades were examined in each area. Examination was by probability plot and by a Monte Carlo simulation process that assessed risk to production from high-grade samples. The analyses determined that about 4.5% of the metal may be at risk due to the very high grade samples. The capping grade, which reduced the risk an appropriate amount for each region, was 30 g Au/t for AC MA and 20 g Au/t for Lewis. This cap was applied to the 2 m composite data prior to grade interpolation.

Variography

The spatial continuity of capped composite data was then examined by variography. The approach used to develop the correlogram model for the gold within the estimation shells is to calculate a relatively large number of sample correlograms in several directions. Directional sample correlograms are calculated along horizontal azimuths of 0°, 30°, 60°, 120°, 150°, 180°, 210°, 240°, 270°, 300° and 330°. For each azimuth,

sample correlograms are also calculated at a dip of 30° and 60°. A final correlogram is calculated in the vertical direction. Base on the 37 sample correlograms, an algorithm determines the best-fit model. The model consists of a nugget-effect, two nested-structure variance contributions; ranges for the variance contributions; and the model type (exponential for all correlogams). After fitting the variance parameters, the algorithm then fits an ellipsoid to the ranges from the directional models for each structure. The anisotropy in grade variation is given by the two ellipsoids.

Gold correlograms are calculated within the PACK-generated shell to reflect the variability in mineralization most likely to be exploited by mining. These correlograms show that the greatest continuity tends to extend to the northeast and be near vertical (see Appendix C).

Density

Bulk density data were taken from the work done by Placer. Placer had compiled an extensive data set for the felsic intrusive rocks and the sediments. A value of 2.65 was assigned to the mineralized intrusive units, based on 1,190 measurements. A value of 2.71 was assigned to mineralized sediments, based on 700 measurements. MRDI supports the use of these numbers. The density values do not vary by much in each unit grouping; this reflects the uniform lithology of the felsic intrusive units and sediments and the low amount of sulphide minerals in these deposits.

Interpolation Plan

Modelling consisted of grade interpolation by ordinary kriging (KG). Both capped and uncapped kriged grades were interpolated. Nearest-neighbour (NN) grades (capped grade only) were also interpolated for validation purposes. The interpolation parameters are listed in Table 17.1. A two-pass approach was used within each PACK shell. The first pass set the grade estimate for material likely to be classified as inferred mineral resource, and the second set values for areas likely to be classified as measured or indicated mineral resource. Only a single interpolation pass was used for material outside the PACK shells.

The interpolation-selected parameters were based on accumulated geological understanding of the gold mineralization as well as variogram analysis. They were also adjusted to incorporate an appropriate amount of grade smoothing into the resource model. The amount of smoothing was checked by change-of-support analysis for each region using the using the Discrete Gaussian or Hermitian polynominal change-of-support method described by Journel and Huijbregts (Mining Geostatistics, Academic Press, 1978). The distribution of hypothetical block grades derived by the Discrete Gaussian change-of-support is compared to the estimated model grade distribution by means of grade-tonnage curves. The grade-tonnage curves allow comparison of the histogram of the two grade distributions in a format familiar to mining. If the estimation procedure has adequately predicted grades for the selected block size, the grade-tonnage curves should match fairly closely. If the curves diverge significantly, there is a problem with the estimated resource.

This method uses the “de-clustered” distribution of composite grades from a nearest-neighbour or polygonal model to predict the distribution of grades in blocks. In this case, the blocks used in the model are 10 m x 10 m x 5 m. The unadjusted polygonal model assumes much more selectivity for ore and waste than is actually possible in mining practice, since each block contains many sample-sized (core sample) volumes that are averaged together. This means that within a block, part of the sample-sized volumes may be ore (above the mining cut-off) and part may be waste. Hence, the distribution of the grade of the blocks is not likely to look like the distribution of grades from composite samples derived from the polygonal estimate. The method assumes that the distribution of the blocks will become more symmetric as the variance of the block distribution is reduced, i.e., as the mining blocks get bigger. The distribution of

hypothetical block grades derived by this method is then compared to the estimated grade distribution to be validated by means of grade-tonnage curves.

The distribution of calculated 10 m x 10 m x 5 m blocks for gold inside the PACK shell for ACMA and Lewis is shown by the dashed lines on the grade-tonnage curve (see Appendix C). This is the distribution of grades based on 10 m blocks obtained from the change-of-support model. The solid line in the figure shows the grade-tonnage distribution obtained from the block estimates. The grade-tonnage prediction produced by the ordinary kriged model shows that grade and tonnage estimates are validated by the change-of-support calculations (grade tonnage prediction curves were validated in the 2.5 to 3.5 g Au/t cut-off range).

Model Validation

MRDI completed a detailed validation of the ACMA and Lewis resource block models. The model was checked for proper coding of drill hole intervals and block model cells, in both section and plan. Coding was found to be properly done. Gold grade interpolation was examined relative to drill hole composite values by inspecting sections and plans. The checks showed good agreement between drill hole composite values and model cell values. Examples of representative sections and plans containing block model gold grades, drill hole composite gold values and PACK shell outlines are included in Appendix D.

MRDI also checked the block model estimates for global bias by comparing the average gold grades (with no cut-off) from the model (ordinary kriging) with means from nearest-neighbour estimates. (The nearest-neighbour estimator produces a theoretically unbiased estimate of the average value when no cut-off grade is imposed and is a good basis for checking the performance of different estimation methods.) Results summarized in Table 17.2 show no evidence of bias in the estimate.

Resource Classification and Summary

The mineralization of the Donlin Creek project as of January 24, 2002, is classified as Measured, Indicated and Inferred Mineral Resources. The classified Mineral Resources are shown in Table 17.3. MRDI and Nova-Gold selected cut-off grades of 3.5 g Au/t, 2.5 g Au/t and 1.5 g Au/t as being representative of the large-scale open pit mining operation that would be potentially economic for gold prices. In this case the prices corresponding to the cut-off grades are U.S.$250, U.S.$300 and U.S.$350 per ounce of gold, respectively.

The logic for mineral resource classification of ACMA and Lewis was consistent with the CIM definitions referred to in National Instrument 43-101. The Indicated mineral resource category is supported by the present drilling grids (nominal 25 m to 35 m) over the ACMA and Lewis deposits. Geologic and grade continuity can be reasonably assumed for this level of resource classification. Blocks containing an estimate where two or more samples were from different holes (one drill hole within 25 m, the other within 35 m) were classified as Indicated mineral resource. To fill in irregular “strips” and “holes” that are a consequence of computer-assigned categories, the sections and plans were inspected and the probability of exceeding the threshold values calculated in the blocks used to help smooth out the assignment process. Honouring the two or more drill hole rule, blocks that had a probability between 0.40 and 0.75 of being above the respective gold grade threshold were also assigned as Indicated mineral resource. This eliminated the odd “strips” and “holes” in the classification.

The Measured mineral resource category is supported only in local areas that have a nominal drill grid spacing of about 15 m. Blocks containing an estimate where two or more samples were from different holes (one drill hole within 10 m, the other within 15 m) were classified as Measured mineral resource. In addition, blocks that had a probability between 0.75 and 1.00 of being above the respective gold grade threshold and that honoured the two drill hole rule as described in the previous paragraph were also assigned as Measured mineral resource.

Estimated blocks not classified as Measured or Indicated mineral resources were classified as Inferred mineral resources if they were within the PACK shells and the closest composite was within 120 m of the block centre.

OTHER DATA AND INFORMATION

No other data or information is relevant for the review of the Donlin Creek project.

REQUIREMENTS FOR TECHNICAL REPORTS ON PRODUCTION AND DEVELOPMENT PROPERTIES

This section is not relevant for this study.

Bondar-Clegg, a commercial laboratory. However, the results can be evaluated together because the Standard Reference Material (SRM), the blank material and the duplicate protocol were the same. The performance of both SRMs was within acceptable limits and showed that overall the assay process was in control for the work done. However, performance of both standards did suffer somewhat in the most recent (2001) batches. Coincident results of both SRMs in some of the batches from that program fell below the 95% confidence limit. These deviations could indicate a lower-grade bias in samples represented by those batches. As a result, the ACMA resource model could be a slightly conservative for material around the cut-off value. The duplicate sample program showed no bias in the assay process; performance was 90% within ±20% for values greater than 0.15 g Au/t. Good reproducibility of the gold values is demonstrated. The blank sample program worked well and demonstrated negligible contamination in the assay process.

REFERENCES

Bundtzen, T.K., and Miller, M.L., 1997, Precious metals associated with Late Cretaceous – Early Tertiary igneous rocks of southwestern Alaska: Economic Geology Monograph 9. p. 242-286.

Decker, J.E., Blodgett, R.B., Box, S.E., Bundtzen, T.K., Clough, J.G., Coonrad, W.L., Gilbert, W.G., Miller, M.L., Murphy, J.M., Robinson, M.S., and Wallace, W.K., 1994, Geology of southwestern Alaska, in Plafker, G., and Berg, H.C., eds., The geology of North America: Geological Society of America, v. G-1,p. 285-310.