Earth Engine下长时间序列监测
巨流河是清代称呼辽河的名字，她是中国七大江河之一，辽宁百姓的母亲河。哑口海位于台湾南端，是鹅銮鼻灯塔下的一泓湾流，据说汹涌海浪冲击到此，声消音灭。
本次将以哑口海周边区域为对象，研究该海岸近30年来的变化趋势。
为求简便，只取了1988，1998，2008和2018的典型无云影像来分析，如果需要精准研究的话就要取很长的时间序列来研究了。具体过程也只涵盖了随机森林分类，植被NDVI计算和建筑区求像元数。其中勾选的训练样区应该在30年内均无变化或为同一地物，如果变化大则需要对每年的影像建立单独的训练样本。
之前想一起把趋势图画了，但现成的ui.Chart.image.byClass()只能对单景影像画图，而ui.Chart()里面的数据又需要为number类型，花了好长时间也没能将计算出的Number Object类型转换为number类型，所以就只有将就用List看一下了，况且科学的图表还是其他软件在行。
计算结果如下图，第一个List表示每年的植被区域NDVI值，第二个List表示每年的建筑区像元数，单像元均为0.09公顷。简单结论：在1988-1998，2008-2018年间该地区为适应旅游规划或生态建设，持续将建筑区规划为绿地；近30年来，绿地的护养情况也得到了提高。（1998的研究影像成像于11月，处于秋末冬初季，略小于1988年的8月或为正常现象）
计算结果
代码如下（第66行为代码开始的地方，前面为研究区和训练区）：
var veg = /* color: #98ff00 */ee.Feature(
ee.Geometry.MultiPoint(
[[120.8596453364878, 21.902058426152273],
[120.8572849925547, 21.913644945654127],
[120.85393759570411, 21.91205235278763],
[120.85831496081641, 21.901062976747287],
[120.85196348986915, 21.90870784980892],
[120.85041853747657, 21.91475975006526],
[120.85792872271827, 21.90504473264743]]),
{
"region": 0,
"system:index": "0"
}),
bld = /* color: #0b4a8b */ee.Feature(
ee.Geometry.MultiPoint(
[[120.8525004546998, 21.903887372630997],
[120.8534016769288, 21.904882802307007],
[120.85129882506112, 21.90540042299038],
[120.85224296263436, 21.902055763858968],
[120.85121299437264, 21.904723534026147],
[120.85494662932138, 21.917026484354032],
[120.8539595764039, 21.91762368790777],
[120.84803725889901, 21.912965433772182],
[120.85001136473397, 21.90719217237422],
[120.85198547056893, 21.905718957860927],
[120.85155631712655, 21.904683716928112]]),
{
"region": 1,
"system:index": "0"
}),
yakouSea = /* color: #ffc82d */ee.Geometry.Polygon(
[[[120.84563635671248, 21.91545161669814],
[120.84649466359724, 21.915133103589817],
[120.8470096477281, 21.9148942182911],
[120.84726713979353, 21.9144164464916],
[120.84765337789167, 21.913938673089387],
[120.84778212392439, 21.91302293625539],
[120.84773920858015, 21.912067378499433],
[120.84782503926863, 21.91095255301553],
[120.84782503926863, 21.908882139683847],
[120.84769629323591, 21.907727473025293],
[120.84799670064558, 21.906453347231466],
[120.84825419271101, 21.90553756227942],
[120.84902238083521, 21.905473842432095],
[120.8495802803103, 21.905075673047698],
[120.84885071945826, 21.904438599718425],
[120.84855031204859, 21.90384134088545],
[120.84885071945826, 21.901890277919467],
[120.84975194168726, 21.900138280249525],
[120.85142564011255, 21.899262273338277],
[120.85318516922632, 21.898067709782673],
[120.85696171951929, 21.897788976845863],
[120.85957955551783, 21.897709338763818],
[120.8622403068606, 21.89834644217444],
[120.86331319046656, 21.899142817433216],
[120.86365651322046, 21.900815190993278],
[120.86365651322046, 21.901930095798022],
[120.86202573013941, 21.9035228018176],
[120.86026620102564, 21.90615072783202],
[120.85940789414087, 21.90822118085846],
[120.8596653862063, 21.911486064880044],
[120.8602232856814, 21.915029573828004],
[120.85893582535425, 21.916900829498537],
[120.85773419571558, 21.91877206058061],
[120.85679005814234, 21.92112101798072]]]);
/*
* Monitering Yakou Sea Area
* on Vegtaion and Buildings
* in 1988 to 2018
*
* Author: Deserts Tsung, Nov 19，2018
* Chengdu University of Information Technology
*
*/
// Variables Could Use Widely
var NDVI = [0,0,0,0], // NDVI of Vegtation in 4 Years
BLDAREA = [0,0,0,0], // Area of Buildings in 4 Years
bands5 = ['B1','B2','B3','B4','B5','B7'], // Bands to be Used in Classification of Landsat-5
bands8 = ['B1','B2','B3','B4','B5','B6','B7'], // Bands to be Used in Classification of Landsat-8
trainroi = ee.FeatureCollection(bld)
.merge(ee.FeatureCollection(veg)); // Training Regions of Classification
// Begin for
for (var i=0;i<3;i++) {
if (i === 0) {
var img80s = ee.Image('LANDSAT/LT05/C01/T1_SR/LT05_117045_19880801').clip(yakouSea);
} else if (i === 1) {
var img80s = ee.Image('LANDSAT/LT05/C01/T1_SR/LT05_117045_19981101').clip(yakouSea);
} else if (i === 2) {
var img80s = ee.Image('LANDSAT/LT05/C01/T1_SR/LT05_117045_20080824').clip(yakouSea);
} // Select one of These Images to Compute
var training80s = img80s.select(bands5).sampleRegions({
collection:trainroi,
properties:['region'],
scale:30
}),
classifier80s = ee.Classifier.randomForest().train({
features:training80s,
classProperty:'region',
inputProperties:bands5
}),
cls80s = img80s.select(bands5).classify(classifier80s); // Classification Above
var mask_veg80s = cls80s.eq(0),
veg80s = img80s.normalizedDifference(['B4', 'B3']).updateMask(mask_veg80s);
NDVI[i] = veg80s.reduceRegion({
reducer:ee.Reducer.mean(),
bestEffort:true,
}).get('nd'); // Vegtation Index and Counts
var mask_bld80s = cls80s.eq(1),
bld80s = img80s.normalizedDifference(['B4', 'B5']).updateMask(mask_bld80s);
BLDAREA[i] = bld80s.reduceRegion({
reducer:ee.Reducer.count(),
bestEffort:true,
}).get('nd'); // Building Counts
//Map.addLayer(cls80s,{min:0,max:1});
// End for
}
// Compute 2018 Alone
var img = ee.Image('LANDSAT/LC08/C01/T1_SR/LC08_117045_20180921').clip(yakouSea),
training = img.select(bands8).sampleRegions({
collection:trainroi,
properties:['region'],
scale:30
}),
classifier = ee.Classifier.randomForest().train({
features:training,
classProperty:'region',
inputProperties:bands8
}),
cls = img.select(bands8).classify(classifier); // Classification Above
var mask_veg = cls.eq(0),
veg = img.normalizedDifference(['B5', 'B4']).updateMask(mask_veg);
NDVI[3] = veg.reduceRegion({
reducer:ee.Reducer.mean(),
bestEffort:true,
}).get('nd'); // Vegtation Index and Counts
var mask_bld = cls.eq(1),
bld = img.normalizedDifference(['B5', 'B6']).updateMask(mask_bld);
BLDAREA[3] = bld.reduceRegion({
reducer:ee.Reducer.count(),
bestEffort:true,
}).get('nd'); // Building Counts
// Add Layers
//Map.addLayer(cls,{min:0,max:1});
// Print Sattistics
print(NDVI,BLDAREA);