Change Detection With Satellite Time-Series Data

For all players of the modern agricultural market (regardless of whether they are farmers, ag traders or insurers) not only is the assessment of land productivity genuinely urgent, but so is constant monitoring of its efficient use.

Farmers can improve yields through tracking growth dynamics, thus developing healthy field trends based on abnormal cases. Insurers can conduct a more effective assessment of land conditions, not here and now, but over the course of a long period of time using historical data. Ag traders are able to schedule logistic spending plans more accurately. Lastly, scientists can monitor the biosphere to track phenological crop distribution, observe global warming, etc.

Time Series Analysis – What Is It About And Why Is It So Effective?

Such a kind of monitoring is carried out so as to detect and analyze the occurring changes to reveal anomalies caused by improper field management. Having the ability to constantly monitor and accumulate miscellaneous data, you can easily determine trends, derive patterns and identify deviations in your areas of interest (AOI), and therefore take timely measures as well as make reliable business decisions.

Say, for example, you were to build a time series graph using historical data at the beginning of the growing season. In this scenario you suddenly find that this year you have an extremely low yield among one of your fields compared with the performance within the same period of time for a number of years. If caught early you will have ample time and opportunity to identify and eliminate the root of the problem, therefore preventing potentially catastrophic crop losses in the future.

However, sorting out tons of images taken over several years, applying the necessary vegetation indices to each of them, and identifying trends from this massive pile of data is not an easy task – so why not entrust this with a specially trained tool? Time series analysis (TSA) can handle this grind for you!

NDVI, NDWI And NDSI – The Perfect Way To Visualize Your Data

To achieve this level of monitoring, various spectral indices are applied to satellite data. For TSA we have:

• NDVI. It is a basic indicator of plant health. Building time series with NDVI allows you, for instance, to increase the accuracy of the N-fertilizers distribution or based on the state of the leaf canopy from year to year, get at hand precise crop statistics, which can help predict yields for the current year. Moreover, it helps to identify and eliminate the root causes of an anomaly for preventing its occurrence in the future. It is very important to detect an abnormal drop of NDVI at the beginning of the season. This can be achieved through a comparison of the present field performance with the same data but for previous years. If such a kind of drawdown happens, immediately set the task for a scout who will identify the cause of a drop; these may be pests, plant diseases, etc. and you will have enough time to act accordingly, so as not to lose the crop.
• NDSI was designed to detect snow and ice cover. Creating time series using NDSI simplifies their mapping, and this matters because in numerous countries snow is perceived to be one of the most devastating natural disasters and have a huge negative impact on agriculture. Snow cover assessment also plays a major role in weather forecasting, hydrological assessment, research related to melting glaciers, and climate change. Snow cover is a lateral water supply in agriculture, that is a big necessity for plants, as well as protection against freezing in winter months for perennials and many other crops.
• NDWI serves to detect the amount of water in soil or canopy. TSA created with the use of NDWI index will show the water content in soil and plants and the fluctuations over long intervals of time. The accumulation of such data will help identify areas that need additional irrigation as well as early detection of water shortages that can prevent plant wilting, drought, and other negative consequences.

In addition, with the use and assistance of these indices you can create maps that will simplify land assessment and solve a wide range of problems. Not only can agricultural issues be assessed, but also deforestation or pre-assessment of damages caused by natural and man-made disasters, weather forecasting, drought prevention, etc.

Time-Series Approaches Currently Available In GIS Tools

TIMESAT Software Package

TIMESAT is a software package for analyzing time-series of satellite sensor data. TIMESAT is developed to investigate the seasonality of satellite time-series data and their relationship with the dynamic properties of vegetation, such as phenology and temporal development. The temporal domain holds important information about short and long-term vegetation changes.

Pros:

• Free of charge

Cons:

• The Desktop app, installation issues, downloading, etc.
• Works with MODIS, low-resolution imagery only
• Works with AVHRR NDVI data only

TIMESAT software package screen.

ASAP

Pros:

• User-friendly interface, easy-to-use
• Free of charge

Cons:

• Works with a small number of satellites, three in particular
• Only NIRSWIRR or NDVI indices may be used
• The ability to receive data starting from 2012 only

ASAP system screen.

Remote Pixel NDVI Series

This project was built to test the capacity of AWS Lambda to do some fast and simple image processing using python and satellite images. A simple tool for building time series analysis using the NDVI spectral index.

Pros:

• Large AOIs for creating time series – 1000 sq. km
• Free of charge

Cons:

• NDVI only
• Landsat 8 and Sentinel 2 only
• The ability to create time series starting from 2013 only
• No atmospheric correction
• The project remained at the demo level

Remote pixel tool screen.

EOSDA LandViewer – On The Cutting Edge Of Time-Saving

To visualize historical data in dynamics and detect changes we developed the Time series analysis feature – an easy-to-use tool that allows the user to build the spatiotemporal graphs with specific indices applied within set AOI. This visualization is based on the data we acquire from each of the seven satellites individually or from all of them simultaneously.

The feature offers the possibility to monitor the dynamics of crop growth within your area of interest easily and without any additional tools, as well as to monitor the effectiveness of land and water resources usage, track environmental changes, study biodiversity within the selected AOIs, forecast weather, pre-assess damages caused by natural disasters, all right from your browser.

How It Works On EOSDA LandViewer

All that’s required is to set (draw or upload) your area of interest (AOI), select the Time series analysis tool, and it will do the work for you.

NB: your max AOI for time series analysis is limited to 200 sq. km

Creating Time Series

The tool offers three indices to choose from: NDVI, NDWI, and NDSI. As seen in the image above, each index allows you to fulfil a specific range of tasks. Graphs can be created for any time intervals from one month and up to ten years by default. In case they do not suit your needs you are able to set custom time periods using the calendar. In addition to the spectral index and the time period, you can also select the data source, but the icing on the cake is the ability to build the graphs using all of the sensors simultaneously!

Currently, data obtained from such satellites as Sentinel-2, Landsat 8, CBERS-4 WFI, CBERS-4 MUX, CBERS-4 PAN10, Landsat 4-5 TM, and Landsat 4-5 MSS is available for TSA. Once all of the parameters are set, apply them to get the detailed analysis of your AOI for any time interval.

Creating Time Series Graph

When a drawdown of any index appears on the graph, you can visualize this plot on the map and check the analytics to find the reason of such index behavior. An equally important point is that the creation of graphs and analytics takes place using only declouded images.

Splitting By Years

In order to visualize field performance for the same period, but for different years, the Splitting by years option is provided. Divided up by years, the curve lets you easily compare index values across 3 to 10 years on the same graph. Compare values based on previous experience to know whether they are in the normal ranges or not. With this new visualization of Time series analysis you will never miss a trend or anomaly.

Split By Years

Data Downloading

You can download the results in the form of a graph or table.

The download in the form of a graph will provide you with a .png file that displays the minimum, maximum, and average values, as well as the standard deviation for the selected spectral index.

Graph downloading

You can download a .csv file, where all of the necessary data will be presented in the form of a table.

Table downloading

Time series In Practice

Use Case 1

Illegal deforestation monitoring

In recent years the problem of illegal logging has become especially relevant for the Ukrainian Carpathians. Let’s conduct a small private investigation. By simply scrolling the map, find the place where the consequences of deforestation are visible to the naked eye and set the area of interest (AOI) within it.

Use case 1. Setting AOI

Using the Time Series Analysis tool (TSA), create the graph using the NDVI vegetation index. This index will display the amount of biomass present in this area for a number of years. The graph allows defining time intervals when major drops of the index occurred. By disabling the smooth curves with no drawdowns, we can determine the year when there was an extreme decrease. Select this year and the next one to be displayed on the graph.

Use case 1. Creating time series

The graph shows that the amount of biomass sharply decreased in 2004 and remained low throughout 2005, but then increased again. This increase was due to the fact that the trampled and cut part of the forest had since ran wild. With a couple of simple steps, one can easily determine the exact time of illegal logging.

Use Case 2

Using Time series analysis in agriculture

Let’s examine the case of creating time-series for a specific field. At this point, we create time series for those years when the same crops grew on the field. Using NDVI we can form certain trends in the development of this crop. Let’s focus on 2017 and 2019. Consider 2017 as a reference year, as this year we had a historically high crop. Note the increasing of the vegetation index in March, April, and May. This is due to the emergence of shoots and the active crop growth in this period.

In the current year we can also observe an increase of NDVI; however, in early July there is a major drawdown. This NDVI drawdown was caused by abnormally high temperatures and serves as a clear alarm. By comparing the pace of crop development over different years, you can promptly spot anomalies, and thus take action. As can be seen from the graph, the farmer managed to rectify the situation using time series analysis and comparing the current NDVI with the one from the years when the yield was great.

Use case 2. Time series of a problem field

This screen shows the NDVI drawdown in 2019 compared to 2017.

Use Case 3

The following use-case considers not only your fields, but neighboring ones. Your fields are seeded with corn. Field monitoring is carried out through the use of satellite NDVI maps. Using these maps, we discovered plots with disparate NDVI values throughout the field.

Use case 3. NDVI map of a problem field

But how can we understand whether such values are the standard ones or these are deviations? And if these are deviations, how major they are? To answer these questions, we can use the Time series analysis tool. Let’s take other cornfields within our district and create graphs of their performance for the last three months. Thus we can draw some trends in the development of cornfields for the whole district. It’s important to take such steps exactly in your district, because these trends are influenced by many factors such as weather conditions, and the weather is about the same for such narrow areas. Once you have created graphs within your district, you can easily save them in the form of a table.

The performance of the field can be conveniently compared with others since this data is in a tabular form. As a result, you can gain insight on whether the NDVI performance of your fields falls under the relative norm or is completely different from district-wide trends.

Use case 3. Comparing results on two fields

Boost Your Analytics With Time-Series Analysis

In summary, this tool provides you with the exceptional ability to carry out constant monitoring of your area of interest so as to accumulate various data and make its further analysis much more accurate. This will allow you determine trends, derive patterns, and identify deviations within your fields. With our new feature, predicting yields will be easier than ever.

For detailed information on Time-series analysis, check EOSDA LandViewer’s user-guide or email us at support@eos.com

Kateryna Sergieieva

Senior Scientist at EOS Data Analytics

Kateryna Sergieieva joined EOS Data Analytics in 2016. She has a Ph.D. in information technologies and a 15-year experience in remote sensing.

Kateryna is a Senior Scientist at EOSDA. Her specialty is the development of technologies for satellite monitoring of natural and artificial landscapes and surface feature change detection. Kateryna is an expert in the analysis of the state of mining areas, agricultural lands, water objects, and other features based on multi-layer spatial data.

Kateryna is an Associate Professor conducting research at the Dnipro University of Technology. She is the author of over 60 scientific papers.