How we test our methodsYöntemlerimizi nasıl test ediyoruzКак мы тестируем методы
Before a method is sold, we test it on cases where the answer is already known and the method could not have seen it. Every test is registered before it runs, and every result is listed here, including the ones that failed.Bir yöntemi satmadan önce, cevabı zaten bilinen ve yöntemin göremediği vakalarda test ediyoruz. Her test çalıştırılmadan önce kaydedilir ve her sonuç, başarısız olanlar dahil, burada listelenir.Прежде чем продавать метод, мы проверяем его на случаях, где ответ уже известен, а метод не мог его видеть. Каждый тест регистрируется до запуска, и каждый результат, включая неудачные, приведён здесь.
Held outDışarıda tutulurОтложено
Test deposits and the whole block of ground around them (1° or 2°) are removed before training.Test yatakları ve çevrelerindeki tüm alan bloğu (1° ya da 2°) eğitimden önce çıkarılır.Тестовые месторождения и весь блок вокруг них (1° или 2°) убираются до обучения.
Registered firstÖnce kayıtСначала регистрация
Design, data, pass rule and the figure that decides are written down, with a file hash, before the run.Tasarım, veri, başarı kuralı ve kararı verecek ölçü, dosya özetiyle birlikte, çalıştırmadan önce yazılır.Дизайн, данные, критерий и решающий показатель записываются с хешем файла до запуска.
Against chanceŞansa karşıПротив случайности
Scores are compared with chance, and from G55 on also with labels shuffled inside the same blocks, with 95% intervals.Puanlar şansla, G55'ten itibaren aynı bloklar içinde karıştırılmış etiketlerle de, %95 aralıklarıyla karşılaştırılır.Результаты сравниваются со случайностью, а начиная с G55 и с метками, перемешанными внутри тех же блоков, с 95% интервалами.
Gated in softwareYazılımla korunurЗащита в коде
A model that fails its test cannot be switched on as validated.Testini geçemeyen model doğrulanmış olarak devreye alınamaz.Модель, не прошедшая тест, не может быть включена как проверенная.
Every registered test and its verdictKayıtlı her test ve sonucuВсе зарегистрированные тесты и их итоги
79tests with resultssonuçlanan testтестов с результатами
34passedgeçtiпройдены
43did not pass, or passed only in partgeçmedi ya da kısmen geçtiне пройдены или пройдены частично
2controls for another testbaşka bir testin kontrolüконтроли для другого теста
Test titles and findings are given in English, as registered. "Partly" means a test with two registered questions passed one of them. A pass is a statement about the test's region and data, not a promise for other ground. The findings number (§) points to the full write-up, which we send on request.Test adları ve bulgular kayıtta olduğu gibi İngilizcedir. "Kısmen", iki kayıtlı sorusu olan bir testin birini geçtiği anlamına gelir. Geçen test, o testin bölgesi ve verisi hakkındadır; başka alanlar için bir söz değildir. Bulgu numarası (§) talep üzerine gönderdiğimiz tam yazıya işaret eder.Названия тестов и выводы приведены на английском, как при регистрации. «Частично» означает, что тест с двумя зарегистрированными вопросами прошёл по одному из них. Пройденный тест говорит о его регионе и данных, а не обещает того же на другой площади. Номер раздела (§) указывает на полное описание, которое мы высылаем по запросу.
| TestTestТест | What was tested · findingNe test edildi · bulguЧто проверялось · вывод | RegisteredKayıtРегистрация | VerdictSonuçИтог |
|---|---|---|---|
| G1 | REE ML map, blind leave-region-out (run 09) Pre-registered blind backtest with train-as-you-infer features · findings §12 | 2026-10-17 | FailGeçmediНе пройден |
| G1-alk | REE model on alkaline-igneous deposits · findings §31 | 2026-10-18 | FailGeçmediНе пройден |
| G1b-EMIT | Hyperspectral alteration vs background · findings §23 | 2026-10-18 | FailGeçmediНе пройден |
| G1c-EMIT | Hyperspectral carbonate confirmation The carbonate signal does not repeat · findings §28 | 2026-10-18 | FailGeçmediНе пройден |
| G2 | Retrospective discovery on the locked test split · findings §16 | 2026-10-18 | PassGeçtiПройден |
| G3 | Time-honest discovery-date test Would the showing ranking have pointed to deposits found later? · findings §27 | 2026-10-18 | PassGeçtiПройден |
| G4 | Famous REE districts held out · findings §20 | 2026-10-18 | FailGeçmediНе пройден |
| G5 | Drill holes (Western Australia) Rankings do not pick the rich drill holes among drilled holes · findings §37 | 2026-10-18 | FailGeçmediНе пройден |
| G6-A | InSAR controls, same frame InSAR controls, phase A: other mines in the same frame · findings §15 | 2026-10-18 | ControlKontrolКонтроль |
| G6-B | InSAR controls, heap-leach frames InSAR controls, phase B: the Çöpler alert is not distinctive among heap-leach mines · findings §24 | 2026-10-18 | ControlKontrolКонтроль |
| G6-pilot | InSAR Çöpler pre-registered alert rule Sentinel-1 InSAR pilot at Çöpler / İliç · findings §13 | 2026-10-17 | PartlyKısmenЧастично |
| G6b | InSAR own-history rule Radar alert relative to each pad's own history · findings §29 | 2026-10-18 | FailGeçmediНе пройден |
| G7 | Supply Watch vs Mountain Pass production · findings §30 | 2026-10-18 | FailGeçmediНе пройден |
| G8 | Leave-one-continent-out geophysics Leave-one-continent-out test on global data · findings §18 | 2026-10-18 | FailGeçmediНе пройден |
| G9 | Known-showing ranking, global · findings §19 | 2026-10-18 | PassGeçtiПройден |
| G10 | Regional models inside a continent · findings §25 | 2026-10-18 | FailGeçmediНе пройден |
| G11 | Switching rule Switching rule, showings where dense, regional geophysics where sparse · findings §26 | 2026-10-18 | FailGeçmediНе пройден |
| G12 | National database leverage: Australia (OZMIN) A national occurrence database adds large leverage · findings §32 | 2026-10-18 | PassGeçtiПройден |
| G13 | National database leverage: British Columbia (MINFILE) BC MINFILE replicates the leverage · findings §33 | 2026-10-18 | PassGeçtiПройден |
| G14 | National database leverage: Brazil (SGB) Brazil's geological survey database; third replication · findings §34 | 2026-10-18 | PassGeçtiПройден |
| G15 | Airborne magnetics 80 m vs global grids (Australia) · findings §35 | 2026-10-18 | PassGeçtiПройден |
| G16 | Airborne magnetics on top of the known-showing ranking (Australia) · findings §36 | 2026-10-18 | PassGeçtiПройден |
| G17 | National database leverage: Sweden (SGU), FODD deposits Sweden's national database; fourth replication of the leverage · findings §38 | 2026-10-18 | PassGeçtiПройден |
| G18 | Existing Turkish deposits, public data only · findings §39 | 2026-10-19 | FailGeçmediНе пройден |
| G20 | Existing Kazakh deposits, public data only · findings §40 | 2026-10-19 | PassGeçtiПройден |
| G21 | Airborne radiometric thorium for rare earths (Australia) · findings §41 | 2026-10-19 | FailGeçmediНе пройден |
| G21b | Thorium for REE on ground without showings (Australia) Thorium on ground without showings; fail for rare earths; gold with airborne data 4x chance · findings §49 | 2026-10-20 | FailGeçmediНе пройден |
| G24 | Analog search on airborne signatures (Australia) Analog search by signature similarity · findings §42 | 2026-10-19 | FailGeçmediНе пройден |
| G25 | Per-commodity model: showings + airborne geophysics (Australia) One model per commodity beats the showing ranking by 11.8 points · findings §43 | 2026-10-20 | PassGeçtiПройден |
| G26 | Model tournament: 8 learning methods vs showings (Australia) Model tournament; five of eight methods beat showings; the neural network does not · findings §46 | 2026-10-20 | PassGeçtiПройден |
| G27 | Satellite + terrain, 8 methods, deposits vs local background (mostly mine disturbance) Satellite + terrain separate deposits from local background, mostly through mine disturbance; pass as registered, not discovery evidence · findings §44 | 2026-10-20 | PassGeçtiПройден |
| G28 | Satellite + terrain on undeveloped deposits (no mine within 5 km) A weak but real satellite signal on undeveloped ground · findings §51 | 2026-10-20 | PassGeçtiПройден |
| G29 | National gravity + 1:1M lithology and faults on top of G25 (Australia) The national geology map adds 6.6 points on top of G25 · findings §45 | 2026-10-20 | PassGeçtiПройден |
| G30 | Best Australian model: PU + derived geology and geophysics PU learning with derived layers reaches 70% but does not significantly beat G29 · findings §50 | 2026-10-20 | FailGeçmediНе пройден |
| G31 | Transfer: Australia-trained model in British Columbia A model trained in Australia does not beat British Columbia's own showings; without showings it is 3x chance · findings §48 | 2026-10-20 | FailGeçmediНе пройден |
| G33 | National sediment geochemistry (NGSA) on top of G29 (Australia) National sediment geochemistry adds nothing on top of the G29 model; alone it is 4.6x chance · findings §47 | 2026-10-20 | FailGeçmediНе пройден |
| G36 | Gold and copper on ground without showings (Australia) Airborne geophysics + geology map rank unexplored ground at 5.6x chance · findings §53 | 2026-10-21 | PassGeçtiПройден |
| G37 | Clustering + likelihood ratio, no showings (Australia) Clustering + likelihood ratio ranks deposits at 4.6x chance without showings · findings §52 | 2026-10-21 | PassGeçtiПройден |
| G39 | Portable layers (Macrostrat + global grids), Türkiye and Kazakhstan Portable layers do not rank deposits in Türkiye or Kazakhstan · findings §55 | 2026-10-21 | FailGeçmediНе пройден |
| G40 | WA candidate cells checked against MINEDEX state records WA candidates hold 16x more state-recorded sites than random no-record ground · findings §56 | 2026-10-22 | PassGeçtiПройден |
| G41 | Replication of G40 in NT, Qld and NSW state records The state-record check replicates in NT, Queensland and NSW · findings §57 | 2026-10-22 | PassGeçtiПройден |
| G42 | State records as extra training labels (G36 model) · findings §58 | 2026-10-22 | PassGeçtiПройден |
| G43 | State-record check in SA, Vic and Tas The state-record check does not replicate in SA, Victoria and Tasmania · findings §59 | 2026-10-22 | FailGeçmediНе пройден |
| G44 | Regional and no-magnetics models, southern gold and copper A regional model does not help in the south, but the all-Australia model already ranks southern deposits at 66%; fail on the primary · findings §60 | 2026-10-22 | FailGeçmediНе пройден |
| G45 | Protected land removed from training negatives Treating protected land as unknown, not barren, adds 2.8 points · findings §61 | 2026-10-22 | PassGeçtiПройден |
| G46 | Ensemble of supervised model and cluster likelihood Averaging in the cluster likelihood makes the ranking worse · findings §62 | 2026-10-22 | FailGeçmediНе пройден |
| G47 | Learner tournament on the G45 setup No other learner beats XGBoost on the current setup · findings §63 | 2026-10-22 | FailGeçmediНе пройден |
| G48 | Potassium, uranium and K/Th radiometrics Potassium, uranium and K/Th do not improve the ranking · findings §64 | 2026-10-22 | FailGeçmediНе пройден |
| G49 | Licence scale (500 m) ranking around undeveloped WA deposits Inside a licence, 500 m geophysics does not find the deposit · findings §65 | 2026-10-22 | FailGeçmediНе пройден |
| G50 | CNN on airborne image patches vs XGBoost A CNN that sees the airborne images matches XGBoost but does not beat it · findings §66 | 2026-10-22 | FailGeçmediНе пройден |
| G51 | Liberia gold from national airborne geophysics and geology Liberia's 1970s national surveys do not rank gold · findings §67 | 2026-10-23 | FailGeçmediНе пройден |
| G52 | Multi-scale magnetics (depth slices) and analytic signal Magnetic depth slices do not improve the ranking · findings §68 | 2026-10-23 | FailGeçmediНе пройден |
| G53 | WA drill results as labels (hits and barren ground) WA drill results as labels do not help either · findings §69 | 2026-10-23 | FailGeçmediНе пройден |
| G54 | Québec: pre-2005 knowledge ranks later Au/Cu finds on new ground · findings §70 | 2026-10-23 | PassGeçtiПройден |
| G55 | Australian result vs block-shuffled labels and an expert rule The Australian result beats an expert rule clearly, and block-shuffled labels narrowly; mostly regional · findings §71 | 2026-10-23 | PassGeçtiПройден |
| G56 | Deep data: GA 3D susceptibility/density inversion and AusLAMP resistivity Deep data (3D inversion, AusLAMP conductivity) do not improve the ranking · findings §72 | 2026-10-23 | FailGeçmediНе пройден |
| G57 | Brazil without airborne data: earlier SGB records rank later Au/Cu records on new ground Brazil without airborne data; with real local skill but no lead over proximity · findings §73 | 2026-10-23 | PassGeçtiПройден |
| G58 | A confidence figure per candidate cell: rank bands and agreement across 25 bootstrap models A confidence figure per candidate cell · findings §74 | 2026-10-27 | PassGeçtiПройден |
| G59 | Inside a licence with the licence holder data: first-pass company geochemistry ranks where follow-up drilling hits (WA) Inside a licence with the holder's own data; the explorer's rule beats chance, our model does not beat the rule · findings §75 | 2026-10-28 | PartlyKısmenЧастично |
| G60 | Africa with public continental layers (geology map, 90 m lithology, global grids) Africa with public continental layers; regional; showings still rank better · findings §76 | 2026-10-30 | PassGeçtiПройден |
| G61 | Türkiye and Kazakhstan with the USGS porphyry labels Türkiye and Kazakhstan with the USGS porphyry labels; pass for the showing ranking; no-showing model still at chance · findings §78 | 2026-10-30 | PassGeçtiПройден |
| G62 | Known-showing ranking on the current catalogue (G9 repeated) The known-showing ranking on today's catalogue; the gain comes from the national databases · findings §77 | 2026-10-30 | PassGeçtiПройден |
| G63 | Quick-scan models retrained on the current catalogue labels · findings §79 | 2026-10-31 | FailGeçmediНе пройден |
| G64 | Tier 2: 1 km cells with fine public layers against the 0.1-degree model (Australia) Tier 2, 1 km cells in Australia; fine public layers put more Au + Cu deposits into the top 1% of area than the 0.1° model, also away from mines · findings §81 | 2026-10-31 | PassGeçtiПройден |
| G65 | Negative evidence and a novelty flag for choosing licence ground (WA) Negative evidence and a novelty flag in WA; Q1 pass (exploration history improves the ranking); Q2 fail (the novelty flag does not mark weaker ranking) · findings §82 | 2026-10-31 | PartlyKısmenЧастично |
| G66 | Tier 4: hole-scale targeting inside WA projects Tier 4, hole-scale targeting inside WA projects; first-pass data barely tell follow-up holes apart · findings §80 | 2026-10-31 | FailGeçmediНе пройден |
| G67 | Our protocol on the GSC/USGS/GA Zn-Pb datacube (Open File 8836) Our protocol on the GSC/USGS/GA zinc-lead datacube; Q2 pass (known occurrences add to the feature model); the published gradient-boosting accuracy does not hold under block folds, weights of evidence does · findings §83 | 2026-10-31 | PassGeçtiПройден |
| G68 | Finland with GTK open data (bedrock map, structures, till; no airborne grids) Finland with GTK open data; Q1 pass (open vector layers rank deposit cells at district scale, without airborne grids); Q2 fail narrowly (model + showings vs showings alone) · findings §84 | 2026-11-01 | PartlyKısmenЧастично |
| G69 | Ontario with OGS open data and the GSC/USGS/GA datacube Ontario with OGS open data and the datacube; Q1 pass (60.6% in the best 10%); Q2 fail (adding the model to known showings lowers the ranking) · findings §85 | 2026-11-01 | PartlyKısmenЧастично |
| G70 | Newfoundland and Labrador, Yukon, Saskatchewan, Spain and Peru with open survey data The Finland/Ontario protocol in five more regions; Q1 pass in 4 of 5 (Newfoundland and Labrador, Yukon, Saskatchewan, Peru), fail in Spain; Q2 pass only in Newfoundland and Labrador · findings §86 | 2026-11-01 | PartlyKısmenЧастично |
| G71 | Five newer techniques against the registered XGBoost on the G70 regions · findings §87 | 2026-11-02 | FailGeçmediНе пройден |
| G72 | Datacube geophysics and lithosphere grids added to the G70 model in three Canadian regions Datacube geophysics and lithosphere grids added to the G70 model in Newfoundland and Labrador, Yukon and Saskatchewan; pass (+7.0 points) · findings §88 | PassGeçtiПройден | |
| G73 | AlphaEarth satellite embeddings and NASA EMIT minerals added to the Peru copper and Türkiye models AlphaEarth satellite embeddings and EMIT minerals added to the Peru copper and Türkiye models; fail (+5.3, not significant); embeddings alone rank deposits above chance · findings §89 | FailGeçmediНе пройден | |
| G74 | G72 rebuilt from open-licence geophysics (NAMAM magnetics, USGS gravity, worms and LAB) G72 rebuilt from open-licence geophysics; fail (+2.6, -3.9 to +8.2); the open grids keep about a third of the datacube's gain · findings §92 | FailGeçmediНе пройден | |
| G75 | Norway, Sweden and Finland with EGDI occurrences and the 1:1M European geology · findings §90 | PartlyKısmenЧастично | |
| G76 | Satellite embeddings alone as the base model in southern Africa Satellite embeddings alone as the base model in southern Africa; pass (65.4%, lower bound 57.7%); likely helped by mines visible from space · findings §91 | PassGeçtiПройден | |
| G77 | Satellite embeddings on deposits with no mine visible from space, southern Africa Satellite embeddings on deposits with no mine visible from space, southern Africa; pass (50.2%, lower bound 40.0%); about 15 points of G76 were mine recognition · findings §93 | PassGeçtiПройден | |
| G78 | Satellite embeddings as the base model in Central Asia Satellite embeddings as the base model in Central Asia; pass (45.3%, lower bound 33.3%); copper and lead-zinc ranked, gold not · findings §94 | PassGeçtiПройден | |
| G79 | Alaska with public-domain USGS data, plus open geophysics and embeddings Alaska with public-domain USGS data; Q1 pass (71.1%); Q2 fail (showings already at 89.8%); all-data model +3.2, not significant · findings §95 | PartlyKısmenЧастично |
Ready, in pilot, and still in validationHazır, pilotta ve doğrulaması sürenlerГотово, в пилоте и ещё на проверке
| CapabilityYetenekВозможность | TestsTestlerТесты | ResultSonuçРезультат | StatusDurumСтатус |
|---|---|---|---|
| Licence coordinatesRuhsat koordinatlarıКоординаты лицензии | ED50 / TUREF in 3° TM and 6° UTM, automated tests3° TM ve 6° UTM'de ED50 / TUREF, otomatik testlerED50 / TUREF в 3° TM и 6° UTM, автоматические тесты | All tests passTüm testler geçiyorВсе тесты пройдены | ReadyHazırГотово |
| Ranking by known mineralisationBilinen cevherleşmeye göre sıralamaРанжирование по известной минерализации | G9, G3, G12–G14, G17, G62 | The best 10% of ground held 49.3% of 16,124 proven deposits outside the US (95%: 46–52%), above chance on every continent; national databases lifted it in all four countries testedEn iyi %10'luk alan ABD dışındaki 16.124 kanıtlanmış yatağın %49,3'ünü içerdi (%95: 46–52); her kıtada şanstan iyi; ulusal veritabanları test edilen dört ülkenin hepsinde yükselttiЛучшие 10% площади содержали 49,3% из 16 124 доказанных месторождений за пределами США (95%: 46–52%), лучше случайного на каждом континенте; национальные базы улучшили результат во всех четырёх странах | ReadyHazırГотово |
| Model on ground without showings (gold, copper)Zuhur olmayan alanda model (altın, bakır)Модель на площади без проявлений (золото, медь) | G36, G42, G45, G54, G57, G60, G68–G70 | Australia 56–69% of deposit cells in the best 10%; Finland, Ontario, Québec, Brazil, Africa, Newfoundland and Labrador, Yukon, Saskatchewan and Peru above chance. Where showings exist they still rank better, so they come firstAvustralya'da yatak hücrelerinin %56–69'u en iyi %10'da; Finlandiya, Ontario, Québec, Brezilya, Afrika, Newfoundland ve Labrador, Yukon, Saskatchewan ve Peru şanstan iyi. Zuhur olan yerde zuhurlar hâlâ daha iyi sıralar, bu yüzden önce gelirАвстралия: 56–69% ячеек с месторождениями в лучших 10%; Финляндия, Онтарио, Квебек, Бразилия, Африка, Ньюфаундленд и Лабрадор, Юкон, Саскачеван и Перу лучше случайного. Где есть проявления, они ранжируют лучше и идут первыми | By regionBölgeye göreПо регионам |
| Screen targetsTarama hedefleriЦели Screen | Ranked ground, geologist review, field pilots with targets locked before visitsSıralı alan, jeolog incelemesi, hedefleri ziyaret öncesi kilitlenen saha pilotlarıРанжированная площадь, проверка геолога, полевые пилоты с целями, зафиксированными до выезда | Field-planning aid; ranking inside a single licence is not yet proven (G49, G59, G66)Saha planlama aracı; tek ruhsat içinde sıralama henüz kanıtlanmadı (G49, G59, G66)Инструмент планирования полевых работ; ранжирование внутри одной лицензии пока не доказано (G49, G59, G66) | Pilot |
| Rare-earth modelNadir toprak modeliМодель редкоземельных металлов | G1, G1-alk, G1b, G1c, G4, G21, G21b | Not above chance, with hyperspectral, alkaline-rock and thorium layers triedHiperspektral, alkali kayaç ve toryum katmanları denendi; şanstan iyi değilНе лучше случайного, хотя испытаны гиперспектральные данные, щелочные породы и торий | Not offeredSunulmuyorНе предлагается |
| Monitor: footprint, encroachment, water, vegetationMonitor: ayak izi, taşma, su, bitki örtüsüMonitor: контур, нарушения границ, вода, растительность | Beylikova 2019–2024, spring scene and persistence ruleBeylikova 2019–2024, bahar görüntüsü ve süreklilik kuralıБейликова 2019–2024, весенний снимок и правило устойчивости | Farmland false swings cut fromTarla kaynaklı yanlış dalgalanmaЛожные колебания от пашни снижены с 231–694 ha → 53–123 ha'ye indi | ReadyHazırГотово |
| Slope-failure warningŞev kayması uyarısıПредупреждение об обрушении | G6-pilot, G6-A, G6-B, G6b | Optical: not detected. Radar fired 163 days before the İliç failure, but the same rule fires at stable heap-leach padsOptik: tespit edilmedi. Radar İliç kazasından 163 gün önce uyardı, ama aynı kural stabil liç sahalarında da uyarıyorОптика: не обнаружено. Радар сработал за 163 дня до аварии в Иличе, но то же правило срабатывает на стабильных площадках кучного выщелачивания | Not offeredSunulmuyorНе предлагается |
| Mine activity index (Supply Watch)Maden faaliyet endeksi (Supply Watch)Индекс активности рудников (Supply Watch) | G7 | Did not track Mountain Pass production; weather still dominatesMountain Pass üretimini izlemedi; hava koşulları hâlâ baskınНе отследил добычу Маунтин-Пасс; погода по-прежнему доминирует | Internal betaİç betaВнутренняя бета |
Two tests that did not pass, and what changedGeçemeyen iki test ve sonrasında değişenlerДва непройденных теста и что изменилось
Blind test of the rare-earth map (G1)Nadir toprak haritasının kör testi (G1)Слепой тест карты редкоземельных металлов (G1)
0.475 vs 0.50
Share of random points in a 6 × 6 km box that the hidden deposit outscores. 0.50 means no skill.Gizlenmiş yatağın 6 × 6 km kutudaki rastgele noktaları geçme oranı. 0,50 beceri yok demektir.Доля случайных точек в квадрате 6 × 6 км, которые скрытое месторождение превосходит. 0,50 означает отсутствие умения.
DepositsYatakМесторождения
30
95% interval%95 aralık95% интервал
0.395–0.557
RegisteredKayıtРегистрация
Before the run, with site list and pass ruleÇalıştırmadan önce, saha listesi ve başarı kuralıylaДо запуска, со списком объектов и критерием
What changed:Değişen:Что изменилось: hyperspectral EMIT (G1b, G1c), fresh alkaline-igneous deposits (G1-alk) and airborne thorium (G21, G21b) were tested next; none lifted rare earths above chance. The map stays experimental and is not sold.ardından hiperspektral EMIT (G1b, G1c), taze alkali magmatik yataklar (G1-alk) ve havadan toryum (G21, G21b) test edildi; hiçbiri nadir toprakları şansın üstüne çıkarmadı. Harita deneysel kalır ve satılmaz.затем проверены гиперспектральные данные EMIT (G1b, G1c), щелочно-магматические месторождения (G1-alk) и аэроторий (G21, G21b); ничто не подняло редкие земли выше случайного. Карта остаётся экспериментальной и не продаётся.
Çöpler / İliç, optical MonitorÇöpler / İliç, optik MonitorЧёплер / Илич, оптический Monitor
+103 ha within noisegürültü içindeв пределах шума
Same-season change in disturbed ground from May 2023 to April 2024, against a seasonal swing of 609–953 ha.Mayıs 2023'ten Nisan 2024'e aynı mevsim bozulmuş alan değişimi; mevsimsel dalgalanma 609–953 ha.Изменение нарушенной площади в тот же сезон с мая 2023 по апрель 2024 при сезонных колебаниях 609–953 га.
Area monitoredİzlenen alanПлощадь наблюдения
2,497 ha
Why missedNeden kaçtıПочему пропущено
Bare material slid onto bare ground; optical classes do not changeÇıplak malzeme çıplak zemine kaydı; optik sınıflar değişmezГолый материал сполз на голый грунт; оптические классы не меняются
RadarRadarРадар
Fired 163 days before, but also at stable pads (G6-B, G6b)163 gün önce uyardı, ama stabil sahalarda da (G6-B, G6b)Сработал за 163 дня, но и на стабильных площадках (G6-B, G6b)
What changed:Değişen:Что изменилось: snow mask, alerts compare the same season only, and Monitor is not sold for slope stability. Monitor is never a life-safety system.kar maskesi eklendi, uyarılar yalnız aynı mevsimi karşılaştırır ve Monitor şev stabilitesi için satılmaz. Monitor hiçbir zaman can güvenliği sistemi değildir.добавлена маска снега, оповещения сравнивают только один сезон, и Monitor не продаётся для контроля устойчивости склонов. Monitor никогда не является системой безопасности жизни.
Registered and plannedKayıtlı ve planlananЗарегистрированные и запланированные
- PThree field pilots in Türkiye: top 5 targets vs. 5 random control points each, lab assays, misses includedTürkiye'de üç saha pilotu: her birinde ilk 5 hedef ve 5 rastgele kontrol noktası, laboratuvar tahlili, ıskalar dahilТри полевых пилота в Турции: по 5 лучших целей и 5 случайных контрольных точек, лабораторные анализы, промахи включеныMar–May 2027Mart–Mayıs 2027март–май 2027
Each registration and results file is kept with its SHA-256. Researchers who want to replicate a test can request the registered design, data sources and code.Her kayıt ve sonuç dosyası SHA-256 özetiyle saklanır. Bir testi tekrarlamak isteyen araştırmacılar kayıtlı tasarımı, veri kaynaklarını ve kodu talep edebilir.Каждый файл регистрации и результатов хранится с хешем SHA-256. Исследователи, желающие повторить тест, могут запросить зарегистрированный дизайн, источники данных и код. Contact usBize ulaşınСвязаться с нами · Proven resultsKanıtlanmış sonuçlarДоказанные результаты