Classifcation Time Series Error in Google Earth Engine?

Question

In the below code I was going to do classification process for time series data but in training data collection returned this error:

code link: https://code.earthengine.google.com/b81fe0a02ece289b3c788a5ccba67ac9

Map.centerObject(table);
Map.addLayer(table);

var start = '2018-01-01';
var end = '2019-01-01';

var persiannData = function(img){
  return img.clip(table)
  .multiply(255 / 2.0).toInt()
  .copyProperties(img,['system:time_start','system:time_end']);
};


var persiann25km = ee.ImageCollection("NOAA/PERSIANN-CDR")
.filterDate(start, end)
.filterBounds(table).map(persiannData);


print('persiann25km',persiann25km);


/// ancillary data

var dem = ee.Image("USGS/GTOPO30")
.clip(table);


var ndvi = function(img){
  return img.clip(table).addBands(dem)
  .copyProperties(img,['system:time_start','system:time_end']);
};

var evi = function(img){
  return img.clip(table)
  .copyProperties(img,['system:time_start','system:time_end']);
};

var lst = function(img){
  return img.clip(table).select('LST_Day_1km').multiply(0.02)
  .copyProperties(img,['system:time_start','system:time_end']);
};

var modisNDVI = ee.ImageCollection("MODIS/MOD09GA_006_NDVI")
.filterDate(start, end)
.filterBounds(table)
.map(ndvi);

var modisEVI = ee.ImageCollection("MODIS/MOD09GA_006_EVI")
.filterDate(start, end)
.filterBounds(table)
.map(evi);

var modisLST = ee.ImageCollection("MODIS/006/MOD11A1")
.filterDate(start, end)
.filterBounds(table)
.map(lst);


// data integration

var modisDataset = modisNDVI.combine(modisEVI).combine(modisLST);

print('modisDataset',modisDataset);



var innerJoin = ee.Join.inner();

var filterTimeEq = ee.Filter.equals({
  leftField:'system:time_start' ,
  rightField: 'system:time_start'
});


var innerJoinModis = innerJoin.apply(modisDataset, persiann25km, filterTimeEq);

print('persiann & modis', innerJoinModis);


var datasetMap = ee.ImageCollection(innerJoinModis.map(function(feature){
  
  return ee.Image.cat(feature.get('primary'), feature.get('secondary'));
  
}));


print('datasetMap',datasetMap);

// modelling process


var classificationProcess = datasetMap.map(function(img){
  
  var bandNames = img.select('NDVI','EVI','elevation','LST_Day_1km')
  .bandNames();
  
  var trainingData = img.stratifiedSample({
    numPoints: 100,
    classBand: 'precipitation',
    region: table,
    scale: 1000,
    });
    
  var classifier = ee.Classifier.smileRandomForest(80)
  .train({
    features: trainingData,
    classProperty: 'precipitation',
    inputProperties: bandNames
  });
  
  var classified = img.select('NDVI','EVI','elevation','LST_Day_1km')
  .classify(classifier);
  
  
  return classified
  .copyProperties(img,['system:time_start','system:time_end']);

});


print('classificationProcess',classificationProcess);

Answer 1

Some of your inputs don't exist on all days. (Specifically, there is no LST on day 42 and others). You can't train a classifier with missing points in your data, so the classifiers for those days fail.

print(ee.Image(datasetMap.toList(1, 42).get(0))
    .reduceRegion(ee.Reducer.count(), table.geometry()))

That said, Creating a new classifier for each day doesn't sound like a good plan. You won't have any way to judge their quality. Some of them will be good and some will be bad (and some wont work at all, like this one). So you wont know if the classification results on any particular day are due to a good/bad classifier or good/bad training from that day.