Earth Engine에서 벡터에서 래스터로 보간하면 FeatureCollection에서 Image이 생성됩니다. 특히 Earth Engine은 지형지물의 속성에 저장된 숫자 데이터를 사용하여 지형지물 외부의 새 위치에서 값을 보간합니다. 보간 결과 지정된 거리까지 보간된 값의 연속적인 Image가 생성됩니다.
역거리 가중치 보간
Earth Engine의 역거리 가중치 (IDW) 함수는 Basso et al. (1999)에 설명된 메서드를 기반으로 합니다. 역거리에 감쇠 계수 (gamma)의 형태로 추가 컨트롤 매개변수가 추가됩니다. 다른 매개변수로는 보간할 속성의 평균 및 표준 편차와 보간할 최대 범위 거리가 있습니다. 다음 예에서는 원래 래스터 데이터 세트의 공간 간격을 메우기 위해 보간된
메탄 농도 표면을 만듭니다. FeatureCollection는 2주간의 메탄 합성물을 샘플링하여 생성됩니다.
range 매개변수에 지정된 대로 가장 가까운 측정소에서 최대 70km까지만 보간이 적용됩니다.
크리깅
Kriging은 반변동의 모델링된 추정치를 사용하여 알려진 위치의 값을 최적으로 조합한 보간 값의 이미지를 만드는 보간 방법입니다.
크리깅 추정기에는 알려진 데이터 포인트에 맞는 반변동 그래프의 모양을 설명하는 매개변수가 필요합니다. 이러한 매개변수는 그림 1에 설명되어 있습니다.
그림 1. 이상화된 변리오그램 함수에 표시된 nugget, sill, range 매개변수
다음 예에서는 임의의 위치에서 해수면 온도 (SST) 이미지를 샘플링한 다음 Kriging을 사용하여 샘플에서 SST를 보간합니다.
[null,null,["최종 업데이트: 2025-07-25(UTC)"],[[["\u003cp\u003eEarth Engine interpolates numeric data from vector features to create continuous raster images.\u003c/p\u003e\n"],["\u003cp\u003eInverse Distance Weighting (IDW) interpolation estimates values based on the distance and decay factor from known data points.\u003c/p\u003e\n"],["\u003cp\u003eKriging utilizes a semi-variogram model to produce an optimal interpolation based on spatial relationships of known values.\u003c/p\u003e\n"],["\u003cp\u003eBoth methods offer customizable parameters to control the interpolation process, like range, maximum distance, and model-specific settings.\u003c/p\u003e\n"]]],["Earth Engine interpolates numeric data from a `FeatureCollection` to create a continuous `Image`. Inverse Distance Weighted (IDW) interpolation uses a decay factor (`gamma`) and distance parameters to estimate values, demonstrated by interpolating methane concentration data. Kriging interpolation, another method, uses semi-variance estimates (`nugget`, `sill`, `range`) to generate interpolated values, exemplified through sea surface temperature interpolation. Both methods sample raster data to create `FeatureCollections` for interpolation. The `maxDistance` parameter determines the interpolation neighborhood's size in Kriging.\n"],null,["# Vector to Raster Interpolation\n\nInterpolation from vector to raster in Earth Engine creates an `Image`\nfrom a `FeatureCollection`. Specifically, Earth Engine uses numeric data\nstored in a property of the features to interpolate values at new locations outside\nof the features. The interpolation results in a continuous `Image` of\ninterpolated values up to the distance specified.\n\nInverse Distance Weighted Interpolation\n---------------------------------------\n\nThe inverse distance weighting (IDW) function in Earth Engine is based on the method\ndescribed by\n[Basso\net al. (1999)](https://ieeexplore.ieee.org/abstract/document/805606). An additional control parameter is added in the form of a\ndecay factor (`gamma`) on the inverse distance. Other parameters include the\nmean and standard deviation of the property to interpolate and the maximum range\ndistance over which to interpolate. The following example creates an interpolated surface of\n[methane concentration](https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_OFFL_L3_CH4) to fill spatial gaps in the original raster dataset. The\n`FeatureCollection` is generated by sampling a two-week methane composite. \n\n```gdscript\n// Import two weeks of S5P methane and composite by mean.\nvar ch4 = ee.ImageCollection('COPERNICUS/S5P/OFFL/L3_CH4')\n .select('CH4_column_volume_mixing_ratio_dry_air')\n .filterDate('2019-08-01', '2019-08-15')\n .mean()\n .rename('ch4');\n\n// Define an area to perform interpolation over.\nvar aoi =\n ee.Geometry.Polygon(\n [[[-95.68487605978851, 43.09844605027055],\n [-95.68487605978851, 37.39358590079781],\n [-87.96148738791351, 37.39358590079781],\n [-87.96148738791351, 43.09844605027055]]], null, false);\n\n// Sample the methane composite to generate a FeatureCollection.\nvar samples = ch4.addBands(ee.Image.pixelLonLat())\n .sample({region: aoi, numPixels: 1500,\n scale:1000, projection: 'EPSG:4326'})\n .map(function(sample) {\n var lat = sample.get('latitude');\n var lon = sample.get('longitude');\n var ch4 = sample.get('ch4');\n return ee.Feature(ee.Geometry.Point([lon, lat]), {ch4: ch4});\n });\n\n// Combine mean and standard deviation reducers for efficiency.\nvar combinedReducer = ee.Reducer.mean().combine({\n reducer2: ee.Reducer.stdDev(),\n sharedInputs: true});\n\n// Estimate global mean and standard deviation from the points.\nvar stats = samples.reduceColumns({\n reducer: combinedReducer,\n selectors: ['ch4']});\n\n// Do the interpolation, valid to 70 kilometers.\nvar interpolated = samples.inverseDistance({\n range: 7e4,\n propertyName: 'ch4',\n mean: stats.get('mean'),\n stdDev: stats.get('stdDev'),\n gamma: 0.3});\n\n// Define visualization arguments.\nvar band_viz = {\n min: 1800,\n max: 1900,\n palette: ['0D0887', '5B02A3', '9A179B', 'CB4678',\n 'EB7852', 'FBB32F', 'F0F921']};\n\n// Display to map.\nMap.centerObject(aoi, 7);\nMap.addLayer(ch4, band_viz, 'CH4');\nMap.addLayer(interpolated, band_viz, 'CH4 Interpolated');\n```\n\nNote that, as specified by the `range` parameter, the interpolation only\nexists up to 70 kilometers from the nearest measurement station.\n\nKriging\n-------\n\n[Kriging](https://en.wikipedia.org/wiki/Kriging) is an interpolation method\nthat uses a modeled estimate of\n[semi-variance](https://en.wikipedia.org/wiki/Semivariance) to create an image\nof interpolated values that is an optimal combination of the values at known locations.\nThe Kriging estimator requires parameters that describe the shape of a\n[semi-variogram](https://en.wikipedia.org/wiki/Variogram) fit to the known data\npoints. These parameters are illustrated by Figure 1.\nFigure 1. The `nugget`, `sill` and `range` parameters illustrated on a idealized variogram function.\n\nThe following example samples a sea surface temperature (SST) image at random locations,\nthen interpolates SST from the sample using Kriging: \n\n```cplint\n// Load an image of sea surface temperature (SST).\nvar sst = ee.Image('NOAA/AVHRR_Pathfinder_V52_L3/20120802025048')\n .select('sea_surface_temperature')\n .rename('sst')\n .divide(100);\n\n// Define a geometry in which to sample points\nvar geometry = ee.Geometry.Rectangle([-65.60, 31.75, -52.18, 43.12]);\n\n// Sample the SST image at 1000 random locations.\nvar samples = sst.addBands(ee.Image.pixelLonLat())\n .sample({region: geometry, numPixels: 1000})\n .map(function(sample) {\n var lat = sample.get('latitude');\n var lon = sample.get('longitude');\n var sst = sample.get('sst');\n return ee.Feature(ee.Geometry.Point([lon, lat]), {sst: sst});\n });\n\n// Interpolate SST from the sampled points.\nvar interpolated = samples.kriging({\n propertyName: 'sst',\n shape: 'exponential',\n range: 100 * 1000,\n sill: 1.0,\n nugget: 0.1,\n maxDistance: 100 * 1000,\n reducer: 'mean',\n});\n\nvar colors = ['00007F', '0000FF', '0074FF',\n '0DFFEA', '8CFF41', 'FFDD00',\n 'FF3700', 'C30000', '790000'];\nvar vis = {min:-3, max:40, palette: colors};\n\nMap.setCenter(-60.029, 36.457, 5);\nMap.addLayer(interpolated, vis, 'Interpolated');\nMap.addLayer(sst, vis, 'Raw SST');\nMap.addLayer(samples, {}, 'Samples', false);\n```\n\nThe size of the neighborhood in which to perform the interpolation is specified by the\n`maxDistance` parameter. Larger sizes will result in smoother output but\nslower computations."]]