Spatio-temporal analysis of cropping intensification in western Morocco
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Title of Series | ||
Number of Parts | 57 | |
Author | ||
License | CC Attribution 3.0 Germany: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
Identifiers | 10.5446/55577 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | |
Genre |
11
32
38
40
50
53
54
57
00:00
Plot (narrative)Temporal logicObservational studyTime evolutionInformation securityMaxima and minimaSystem programmingSoftware developerMathematical analysisGoogolComputer-generated imageryPoint cloudStatisticsConsistencyInterpolationSpline (mathematics)AlgorithmFunction (mathematics)OrthogonalityLinear mapCluster samplingPopulation densityObservational studyDynamical systemCovering spaceComputing platformFrequencySatelliteStrategy gameMultiplication signInformation securityRow (database)MappingMereologyFlow separationAreaMedical imagingDatabaseInformationPoint cloudStapeldateiLengthAuditory maskingEvoluteAcoustic shadowAlgorithmScripting languageEntire functionGene clusterPopulation densityFunktionalanalysisLinearizationReduction of orderDemoscenePixelMathematical analysisSparse matrixTime seriesDescriptive statistics2 (number)Computer animation
07:58
Cluster samplingTemporal logicDistribution (mathematics)Population densityAreaTime seriesAreaPerformance appraisalComputer animation
09:00
Cluster samplingAreaNormed vector spaceData fusionPopulation densityData modelSoftware testingAreaMedical imagingGene clusterDistribution (mathematics)Multiplication signCovering spaceTime seriesSurfaceMappingWater vaporEndliche ModelltheorieMereologyTime zoneoutputSound effectSpacetimePopulation densityResultantTemporal logicSparse matrixEvoluteChannel capacityCASE <Informatik>SatellitePoint (geometry)Software testingMaizeLocal ringComputer animation
12:53
MappingFocus (optics)Process (computing)MeasurementType theoryMedical imagingAreaPhysical systemCase moddingNumberPoint (geometry)Gene clusterMeeting/Interview
Transcript: English(auto-generated)
00:10
Good morning, everyone. Thank you for giving me again this opportunity to present my work in this workshop. My study is about spatio-temporal analysis of crop intensification in western of Morocco.
00:32
In this presentation, I will start by giving an overview about the study. Then I will present briefly the study area, the user data and methods to carry out this study.
00:45
Later, I will present the findings and finally, I will sum up by giving the main conclusions and the future work because this study is not completely finished. Cropping intensification is defined as cultivating a plot of land several times a year to obtain maximum profit.
01:07
It is a strategy operated for the profitability of equipment and the hydro agricultural facilities and contributes to the country's food security. The use of earth observation data has shown great interest in
01:25
studying agricultural systems, especially with well-developed technological progress in recent years. This study aims to measure the spatio-temporal evolution of cropping intensification through analysis of vegetation,
01:44
cover cropping, dynamic use in satellite images, the Google Earth in giant platforms and R packages. The study area, which is the city district, is located in the western of Morocco and is characterized by a semi-arid climate.
02:07
It covers a large part of the irrigated scheme where irrigation has been implemented progressively on different years. We used Landsat images of two paths and two rows that overlap on the study area.
02:29
We performed the data preparation using Google Earth in giant following three steps. The first one is harmonizing Landsat images by transforming TM and OTM plus to OLS surface reflectance by using OLS regression.
02:48
The second step is cloud and shadow masking by applying the CF mask algorithm.
03:00
And then we calculated the NDVI to build our time series. In the study area, cropping season starts in September of the first years. And ends in August of the following year.
03:24
After developing and testing the script in Google Earth in giant for data preparation, we retrieved the information about available images from 1985 to 2020 by filtering images containing less than 50 percent of clouds.
03:43
And by viewing this data using a blot, it appears that the database for 1998 cannot be used to build a consistent time series. In this regard, we limited this work to start the time series from September 1998 to August 2020.
04:04
After time series length definition, we ran a batch download of each NDVI image of the last cropping season to Google Drive. Then we tested different aggregation to decide which one contains least missing pixels and is convenient to our needs.
04:24
To build our time series, we started by mosaicing image of the same date from several scenes covering the study area. Then we made an aggregation of the images by week and by month, which ended up to select.
04:47
And then we came back again to Google Earth in giant to reduce image collection by month for the entire time series and batch the download to the Google Drive again.
05:01
After this batch reducing images by month, there still be some missing pixels. And to deal with that, we tested four different gap filling methods, linear spline, Stein and data interpolating empirical orthogonal functions.
05:23
Visually, there is no significant difference between methods, although we have selected the Stein method according to its algorithm logic. Before starting the data clustering, we selected only the vegetated areas along the cropping season by applying a pixel mask.
05:44
And the non-vegetated area is a signet to the seventh cluster later. Then we selected the optimal numbers of clusters for coming algorithm, which is determined at six.
06:00
After clustering, we classified cluster by irrigated area and the run feed area, and then we gave them meaningful descriptions. The first cluster is an annual very dense vegetation cover. This cluster represents the area covered by vegetation all the year with an NDVI greater than 0.35 during
06:29
all month and an NDVI greater than or equal to 0.6 during seven months from November to May. The second cluster is less vegetated area characterized by a very sparse vegetation cover with an NDVI that does not exceed 0.35 throughout the year.
06:55
The cluster three is a dense late vegetation with an NDVI between 0.6 and 0.7.
07:02
The cluster four is a very dense vegetation in mid-season with an NDVI between 0.7 and 0.8. The cluster five is medium late vegetation with an NDVI around 0.6 during the period from January to March.
07:21
The last cluster six is medium early vegetation with an NDVI between 0.5 and 0.6 during the period from December to February.
07:41
The cluster six appears in several crop in season maps, especially in run feed data. A non-vegetated area appears clearly in some maps and the irrigated area is perfectly discriminated in all maps. To evaluate cluster spatiotemporal distribution, we selected the foremost contrasted and representative years of the time series.
08:11
The cluster one is mainly located in the irrigated area, while the cluster two is located on the run feed area.
08:27
The two clusters, two and seven, seven is non-vegetated area and two is less vegetated area, are located especially in run feed area.
08:47
The cluster seven, which is a non-vegetated area, is located in sandy soils with wind deposition in fissiletic soils and in built-up areas.
09:04
The two clusters three and four are located in irrigated area, but the cluster four is present in run feed area in some crop in season, which is related to sufficient precipitation. The two clusters six and five are located mainly in run feed area.
09:32
To evaluate the relationship between clusters spatiotemporal distribution and precipitation, we selected three different kinds of situation.
09:42
The first one is characterized by water deficit and poor annual rainfall distribution. We can say it is a drought year. We observe that the non-vegetated area covers a large surface in run feed zones and also the certain part of the irrigated area it was not equipped during the first crop in seasons.
10:05
The affected areas are generally sandy soils with low water retention capacity and planted by sparse perennial vegetation such as vineyards or very low density crops such as local corn and legumes.
10:24
The non-vegetated area extends into adjacent soils during drought years. The second case is characterized by water sufficiency and poor annual rainfall distribution, where cluster six is the most dominant in the run feed area.
10:45
And the third and last case is characterized by the presence of sufficient water with a good distribution of rainfall in the year. We can say it is a run year. We can conclude by the following point. The satellite image time series can be very effective in measuring crop intensification.
11:09
The introduction of irrigation has been accompanied by a substantial intensification of agricultural activities, which is reflected by the presence of a dense vegetation cover during the whole cropping season.
11:23
In the two decades of time series, the agricultural space was disturbed by climate change effects with irregularity regularly and low rainfall during the first decades and until 19 runny and dry years during the last decades.
11:44
The temporal NDVI evolution explains strongly this situation in both run feed and irrigated areas. The results also revealed that during the last three cropping seasons, the NDVI dropped in the irrigated area due to insufficient irrigation water in the dam.
12:07
Our future work will include the following topics among others. Enhanced time series density by using maybe MODIS data with the fusion and we include the last cropping season 2021.
12:21
Test orders, clustering, classification and modeling methods use available grown through data of some cropping seasons to perform the crop mapping. I'm looking forward to your comments and recommendations to input some quality ideas in the future work. Thank you.
12:52
Thank you, Abdulkrim. Let's see if there's some questions. Just chatting the chat also here in the room. Any questions for Abdulkrim?
13:07
So your focus was on mapping cropping intensification, but not, for example, crop types or yields. Is that correct? Because you don't have any measurements going back.
13:22
We don't understand yet the truth of data. We want only to measure the presence of crops on lands during the year. In these clusters, you pick up six clusters, but how did you come to that number of six clusters?
13:46
Did you try detecting objectively number of clusters or how did you come to six clusters? Why six clusters? So we have tested many attempts and we find that this number can give us a good point to describe the situation on the study area.
14:13
Okay, there's one more question I have for you. Like you said that you use the Google Earth Engine to pre
14:21
-process the Landsat images, then you downloaded some images, then you uploaded back to Google Earth Engine, if I understand correctly. I prepared all the images on Google Earth in China, then when I downloaded the collection
14:42
of mods for each cropping season to Google Drive, and I started the processing in R packages. Okay, I'm just wondering, you know, how efficient is that to pass the data, you know, from
15:02
one system to the other. So there was a lot of download and upload. Is that correct? There was a lot of upload and download, you have a lot of data so you had to upload to Google Cloud, then download. Is that correct or how much of the processing is it?
15:23
Yes, just download from Google Earth.
Recommendations
Series of 4 media