Definition
Satellite remote sensing uses sensors on an artificial satellite to collect electromagnetic radiation or reflection from the surface and process it into images and thematic products. It is a branch of remote sensing. Coverage is wide, the revisit is short, and the record is objective and continuous. It is used for environmental monitoring, pollution tracing, precision agriculture, land use, and disaster assessment.
Types and data sources of satellite remote sensing
By sensor, it includes optical remote sensing (visible, multispectral, and hyperspectral), radar remote sensing (SAR), and thermal infrared. Common sources are Landsat, Sentinel, the Gaofen series, and Jilin-1. Spatial resolution runs from the meter level to the sub-meter level, and revisit runs from several days to real time. Optical images can yield a vegetation index such as NDVI, land-surface temperature, and the extent of water. SAR works in all weather and through cloud, which suits floods and ground deformation.
Applications in Precision Agriculture and Environmental Protection
In precision agriculture, satellite images can produce maps of crop condition, soil moisture, and pest and disease stress, and those maps guide variable work by smart agricultural machinery and fertigation. In environmental protection, with pollution tracing, they can watch discharge outlets, dust, and straw burning. A cloud platform processes the data, and machine learning can interpret it and extract themes automatically. SimtoLink can support intake of remote-sensing data and API integration.
Collaboration with Ground Sensors and UAVs
Satellite images, ground sensors for soil moisture and temperature and humidity, and UAV remote sensing form one monitoring system across space, air, and ground. The satellite covers a wide area on a repeat cycle. The UAV gives higher resolution and can be sent where it is needed. Ground sensors give a continuous reading at a point. Fusing the three improves traceability and the decision. Choose the source for the job, the resolution, how soon the image is needed, and the cost.
The Usable Window of Orbital Data
Satellite remote sensing is what an instrument on the satellite received, as surface reflection or its own emission, at one pass, then turned by geometric and radiometric processing into an image or an index. It describes conditions across an area at that moment, not the continuous reading of one probe in the field. Optical imagery is blocked by cloud. Radar can see through cloud but records backscatter. The two cannot be filled in as the same kind of "greenness." The revisit period is when the satellite comes back, not when the farm needs an answer.
Before using it, write down whether this resolution can see the question. If a field is smaller than one pixel, the pixel mixes road, ditch, and the neighboring field, and the index represents the mixture. Irrigation that needs a decision the same day may already have missed the farming window by the next pass. Mosaics of several scenes must share a projection and an acquisition date. Laying maps from different seasons together and comparing growth treats the seasonal difference as damage.
It is easy to treat an attractive true-color picture as an analysis. A browse image without calibration cannot yield a comparable index. Cloud shadow counted as water or bare soil spreads the contaminated area along the cloud edge. A single-day image used to judge whether an organic certification plot has been replanted does not see input records during the conversion period. Imagery can help check whether the boundary is still there. It cannot replace the field ledger. Comparing brightness across two years without atmospheric correction describes sensor aging or solar elevation as a worse crop.
When paired with a ground point, the point should fall inside the pixel and the land cover should be uniform. A point at the headland and a pixel in the middle of the field that do not match cannot "calibrate" each other. In a rainy week when both radar and optical data are missing, state that the week has no satellite observation, rather than copying the previous scene forward.
Resolution has to be read together with field size. If one pixel contains crop, road, and water at once, the index is an average of the mixture and does not represent the crop itself. Drawing that mixed pixel into a fertilizer prescription fertilizes the road and treats the water as poor growth. Pixels that are clearly not cropland should be removed before discussing variation inside the field. If they cannot be removed, it is better not to use the scene than to explain the mixed value as a lack of fertilizer.
A time series is more useful than a single scene, but the dates must be real. Filling today's cloud with last ten-day period's cloud-free image treats growth from ten days ago as today. Harvest, flooding, or fire inside a revisit period may not be visible until the next orbit. During that gap it cannot be claimed that the satellite is monitoring continuously. At too large an off-nadir angle, features are pulled oblique, and both area and boundary change. That scene should not be used to measure the same field against an orthophoto.
Product level has to be checked as well. A browse image, reflectance after atmospheric correction, and a thematic map someone else has already classified are not the same kind of data. An index calculated on a browse image cannot be compared with a reflectance product. If a classification map reuses class names from another region, the local crop may be labeled as the wrong type. Before use it should be possible to state the version, the pass date, and the processing steps. Without those three, the image is fit to look at, not to calculate with.
Comparing several scenes of the same field also requires a consistent cloud mask. One scene removes thin cloud and another treats thin cloud as crop, and the difference shows false growth change along the cloud edge. The mask version should be stored with the product. If the share of the field covered by cloud is beyond what can be explained, do not issue a prescription for that period. Mark it as no valid observation. Filling this field's index from the neighboring field assigns someone else's crop condition to one's own ground.