IMPUTATION OF REMOTE SENSING TIME SERIES FOR LOW-LATENCY AGRICULTURAL APPLICATIONS
Imputation of remote sensing time series for low-latency agricultural applications is provided. In various embodiments, a first time series of raster data is read. The first time series spans a geographic region and has a first resolution and a first frequency. A second time series of raster data is read. The second time series spans the geographic region and has a second resolution and a second frequency. The second resolution is lower than the first resolution. The second frequency is higher than the first frequency. A mean time series is determined from the first time series of raster data. A predicted time series of values for a location within the geographic region is determined at the first resolution by determining a first time series of values for the location from the first time series of raster data, determining a second time series of values of the location from the second time series of raster data, and determining the predicted time series by multiple linear regression with the first time series dependent on the mean time series and the second time series.
1 . A method comprising:
reading a first time series of raster data, the first time series spanning a geographic region and having a first resolution and a first frequency;
reading a second time series of raster data, the second time series spanning the geographic region and having a second resolution and a second frequency, the second resolution being lower than the first resolution, and the second frequency being higher than the first frequency;
determining a mean time series from the first time series of raster data; and
determining a predicted time series of values for a location within the geographic region at the first resolution by
determining a first time series of values for the location from the first time series of raster data,
determining a second time series of values of the location from the second time series of raster data, and
determining the predicted time series by multiple linear regression with the first time series dependent on the mean time series and the second time series.
2 . The method of claim 1 , further comprising:
smoothing the mean time series.
3 . The method of claim 1 , further comprising:
smoothing the predicted time series of values.
4 . The method of claim 1 , wherein each of the first time series of raster data, the second time series of raster data, the mean time series, and the predicted time series correspond to an agricultural index.
5 . The method of claim 4 , wherein the agricultural index comprises normalized difference vegetation index, land surface water index, or and mean brightness.
6 . The method of claim 1 , further comprising:
reading a crop type mask, wherein
determining the mean time series comprises masking the first time series of raster data according to the crop type mask.
7 . The method of claim 1 , further comprising:
reading a plurality of crop type masks; and
determining a plurality of mean time series from the first time series of raster data, each mean time series corresponding to one of the plurality of crop type masks,
wherein determining each mean time series comprises masking the first time series of raster data according to the respective crop type mask.
8 . The method of claim 7 , wherein determining the mean time series comprises selecting the mean time series from the plurality of mean time series.
9 . The method of claim 7 , wherein determining the mean time series comprises selecting one of the plurality of mean time series most similar to the first time series of values.
10 . The method of claim 9 , further comprising:
determining a crop type associated with the selected one of the plurality of mean time series.
11 . The method of claim 1 , further comprising:
applying a time shift to the mean time series based on the first time series.
12 . The method of claim 11 , further comprising:
determining the time shift by cross-correlation between the mean time series and the first time series.
13 . The method of claim 1 , further comprising:
rescaling the second time series.
14 . The method of claim 13 , wherein rescaling the second time series comprises CDF matching.
15 . A system comprising:
a computing node comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by
a processor of the computing node to cause the processor to perform a method comprising:
reading a first time series of raster data, the first time series spanning a geographic region and having a first resolution and a first frequency;
reading a second time series of raster data, the second time series spanning the geographic region and having a second resolution and a second frequency, the second resolution being lower than the first resolution, and the second frequency being higher than the first frequency;
determining a mean time series from the first time series of raster data; and
determining a predicted time series of values for a location within the geographic region at the first resolution by
determining a first time series of values for the location from the first time series of raster data,
determining a second time series of values of the location from the second time series of raster data, and
determining the predicted time series by multiple linear regression with the first time series dependent on the mean time series and the second time series.
16 . A computer program product for agricultural index prediction, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:
reading a first time series of raster data, the first time series spanning a geographic region and having a first resolution and a first frequency;
reading a second time series of raster data, the second time series spanning the geographic region and having a second resolution and a second frequency, the second resolution being lower than the first resolution, and the second frequency being higher than the first frequency;
determining a mean time series from the first time series of raster data; and
determining a predicted time series of values for a location within the geographic region at the first resolution by
determining a first time series of values for the location from the first time series of raster data,
determining a second time series of values of the location from the second time series of raster data, and
determining the predicted time series by multiple linear regression with the first time series dependent on the mean time series and the second time series.
17 . The computer program product of claim 16 , the method further comprising:
reading a plurality of crop type masks; and
determining a plurality of mean time series from the first time series of raster data, each mean time series corresponding to one of the plurality of crop type masks,
wherein determining each mean time series comprises masking the first time series of raster data according to the respective crop type mask.
18 . The computer program product of claim 17 , wherein determining the mean time series comprises selecting the mean time series from the plurality of mean time series.
19 . The computer program product of claim 17 , wherein determining the mean time series comprises selecting one of the plurality of mean time series most similar to the first time series of values.
20 . The computer program product of claim 19 , the method further comprising:
determining a crop type associated with the selected one of the plurality of mean time series.